CantusManualBack to the site

Cantus — quick-start manual

For Cantus 0.1.0-beta.1 and later, on Windows, macOS and Linux · Leer en español

Cantus is a free, open-source digital audio workstation: record, compose, mix and master a song, from the first note to a finished file. This manual gets you there in an evening. Everything it names is in the program as it is today; where the beta still has gaps, chapter 14 says so.

Cantus at start: the arrangement, the mixer below, the inspector at the right

Contents

  1. Install
  2. First run: set up your audio
  3. The workspace
  4. Record audio and MIDI
  5. Edit: clips, piano roll, automation
  6. Mix: the mixer, inserts and sends
  7. Master: the master chain, Fervor, Latitudo, Culmen
  8. Export
  9. Sounds: the browser and sound packs
  10. Instruments and effects: the device picker
  11. Projects, autosave and settings
  12. Keyboard shortcuts
  13. Privacy and updates
  14. Known issues and limitations
  15. Report a bug

Also: What an AI agent can do in Cantus — the tools, the unit catalogue, the music linter.

Also: Separating stems — a song into drums, bass, vocals and the rest, on your computer. Also: Finding and hearing presets — by style and role, heard before you choose.

Also: Exporting a dancing video — your dancers with the song, for social media or a video editor.

Also: Converting audio to MIDI — a melody, chords or a drum loop into notes on a new track.

Also: Pulsus III: pads on their own strips, a chain per pad, kits from anywhere — split a kit to strips, effects on one pad, a folder or an SFZ as a kit.

Also: The arrangement: time, sections and ripple — insert and delete time on every track, drag a section to move its bars, the ripple mode.

Also: Chorda: recorded pianos and the piano's mixer — the modelled piano or a recorded one (the Salamander, bitKlavier, your own SFZ), its microphones on strips, Add our room, a mic's delay.

Also: Tympanum, the acoustic drummer — modelled and recorded drum kits in their room, their microphones on strips, the bleed, the hybrid voice.

Also: The audio editor — cut, silence, fade, normalise, take the noise out, remove silence and see the spectrum, on a clip's take.

Also: Saltator, a character that dances to your song — the dancer's window, its moves, React to; Fuga the ninja and Volta the mecha lead the cast.

Also: The stage — a dancer over the workspace, on the Stage tab, in a companion window and on stream (OBS).

Also: The recording mode — render every strip that feeds the master so the buffer can go lower while you record.

Also: VCA faders and mixer snapshots — move a group of strips together, store the desk and recall it, cue it on the timeline.

Also: External gear and control surfaces - a hardware synth or effect on a track with its latency measured, and Mackie Control / HUI surfaces.

Also: Importing a character — a dancer wearing your own Lottie, sprite sheet or DragonBones character. Also: Editing a clip with an ARA plug-in — an ARA 2 plug-in that sees the whole recording renders your audio clip.

AI in this manual: AI agents (MCP) — connect Claude or another assistant; The mastering assistant; Listening for AI agents.

1. Install

Cantus runs on Windows, macOS and Linux; every release has one file for each, on the releases page and behind the download button of cantusdaw.com, which gives you your system's:

SHA256SUMS.txt beside them lets you check a download: Get-FileHash <file> on Windows, shasum -a 256 <file> on a Mac, sha256sum <file> on Linux. About 200 MB of disk. Setting up your audio on each system is chapter 2.

On Windows

You need Windows 11, 64-bit (Windows 10 untested).

  1. Download Cantus-0.1.0-beta.1-windows-x64-setup.exe from the releases page.
  2. Run it. It installs for you alone, in %LOCALAPPDATA%\Programs\Cantus, and asks for no administrator rights. This beta is not code-signed yet, so Windows may show "Windows protected your PC": choose More info ▸ Run anyway.
  3. Choose the components: the program (required), Open .cantus projects with Cantus (on), a desktop shortcut (off), the CLAP and VST3 plug-ins for other DAWs (on), and FlexASIO driver (downloaded now) (off — tick it if your sound card has no ASIO driver of its own; see chapter 2).
  4. The Privacy page says what Cantus sends over the network (only a daily question about new versions, and the downloads you ask for). Check for updates once a day (recommended) is ticked; untick it to turn the check off (also later in Settings ▸ Updates).

Portable copy. Unzip Cantus-0.1.0-beta.1-windows-x64-portable.zip anywhere (a USB stick works). Because of the portable.txt beside cantus.exe, that copy keeps its settings in the data folder next to it and writes nothing to your profile.

Uninstall from Windows' Apps. Your settings, presets and autosaves — and the sample cache (up to 16 GB, see chapter 11) — are removed only if you tick that box; your projects are never touched.

On a Mac

  1. Download Cantus-<version>-macos-universal.dmg — one disk image for Apple silicon and Intel Macs.
  2. Open it and drag Cantus onto Applications. Eject the disk image.
  3. Open Cantus from Applications (or Launchpad). A notarised release opens straight away; one that is not yet notarised makes macOS say it cannot verify the developer: right-click Cantus ▸ Open ▸ Open once (or System Settings ▸ Privacy & Security ▸ Open Anyway), and it opens normally from then on.
  4. The first time you record, macOS asks whether Cantus may use the microphone: allow it (later in System Settings ▸ Privacy & Security ▸ Microphone).

Cantus keeps its settings, presets and autosaves in ~/Library/Application Support/Cantus. Uninstall by moving Cantus from Applications to the Bin; that folder is yours to delete or keep.

On Linux

  1. Download Cantus-<version>-linux-x86_64.AppImage — one file that runs on most distributions from the last few years, with nothing to install and no administrator.
  2. Make it executable and open it: in a file manager, Properties ▸ Permissions ▸ Allow executing as a program, or in a terminal chmod +x Cantus-*.AppImage && ./Cantus-*.AppImage. Keep it where you like (a folder such as ~/Applications is a good home): Cantus updates that file in place.
  3. If it does not start and says FUSE is missing, install your distribution's libfuse2 (Ubuntu 22.04: sudo apt install libfuse2; 24.04: libfuse2t64).

Cantus keeps its settings in ~/.config/cantus and its presets, autosaves and sounds in ~/.local/share/cantus (or wherever $XDG_CONFIG_HOME and $XDG_DATA_HOME say). Uninstall by deleting the AppImage. A Cantus from your distribution's packages or from Flatpak is installed and updated by them instead.

2. First run: set up your audio

The first time Cantus starts, the window Set up your audio opens over the music.

The first-run audio setup

  1. Choose your interface. Each audio device has a card: an Audio interface, This computer, Headphones… Click the one you play and listen through. Cantus picks a driver and a buffer for it: ASIO at 128 samples for an interface with its own ASIO driver, otherwise WASAPI (Windows' own audio) at 256.
  2. Listen. Click Test tone: you should hear a soft tone on outputs 1–2.
  3. Play or speak — the meter should move. Strum, sing or speak into what is plugged in; the input that moves is the one you will record from.
  4. Latency. Cantus says how long the sound takes from string to speaker, and whether that is tight enough to play through. For the most accurate takes, cable output 1 to input 1 and click Measure latency; Cantus then places every take to the sample.

Click Record on … and Cantus adds an audio track on that input, armed and ready: press R to record. Skip closes the window; it does not open by itself again (the Setup assistant… button in the audio chip brings it back).

On a Mac the driver is Core Audio, which every interface uses: there is no ASIO to install. Choose the buffer (128 samples is a good start) and the rate on the device page; an aggregate device made in Audio MIDI Setup (two interfaces as one) is listed like any other. The first time you record, macOS asks whether Cantus may use the microphone — allow it, or no input reaches Cantus (System Settings ▸ Privacy & Security ▸ Microphone). Cantus's audio threads work with Core Audio's own as one group, so Apple silicon runs them on its fast cores.

The audio chip

The chip in the top bar (or in its ⋯ menu when the window is narrow) shows the device, driver, rate, buffer and latency — for example UMC204HD · ASIO · 48 k · 128 · 6.1 ms. Its lamp is green when the audio runs, amber after a dropout, red when the device is lost or the driver stopped. Click it to change anything; every change applies at once and is remembered for that interface:

The audio chip's popover

Which driver? Use your interface's own ASIO driver if it has one (Behringer, Focusrite, PreSonus, Steinberg, MOTU, RME…). For a laptop's built-in sound, use WASAPI, or install FlexASIO for lower latency. Get ASIO… in the popover opens a window with FlexASIO's download page, ASIO4ALL's page and the makers' driver pages; after installing a driver, click Rescan drivers. Cantus never installs drivers by itself.

Plugging an interface in offers to switch to it; unplugging one offers your computer's audio, and Cantus goes back to it when it returns. If a driver stops responding, Cantus restarts the audio once by itself. You cannot change the device while recording.

3. The workspace

Cantus opens on a song — the second drop of the demo Neon Tide, looping — so the first thing you hear is music. Press Space to play it. Nothing dances at start: the whole demo (File ▸ Demo songs) keeps its dancer, and you can add one to any song (Saltator).

Every area can be shown or hidden with its window key or its function key (F5 arrangement, Alt+F5 clip browser, F6 channel rack, F7 piano roll, F9 mixer, Alt+F8 browser, I inspector). Drag the splitters to resize (a right-click resets one, a double-click folds or unfolds it); right-click a tab for its menu:

Ctrl+F6 and Ctrl+Shift+F6 move the keyboard's focus from area to area (the focused tab's glyph is lit); Shift+F10 opens what a right-click on the selection opens; Esc closes what opened last. Every area that holds more than it shows scrolls. Save the arrangement of the window under View ▸ Layout ▸ Save layout…; the layout is kept between sessions.

The look. Cantus is drawn in Slate Lapis by default: Slate's calm room, where panels are set apart by value rather than by lines, in a deep night blue; one lapis accent means on or selected and nothing else, and the logo's gold marks the clock, the rulers' numbers, the playhead and the loop. Your track colours are a stripe and a tint until you select a track — then it lights across the window (its header, clips, strip and the inspector). The type is Barlow Semi Condensed, whose figures keep their width as they count; keys and fields are flat wells with a dark edge and a light lip; in silence nothing reads "−∞". Settings ▸ Appearance (Ctrl+,) holds every theme as a picture, in four groups: round 39's families — Slate Lapis (the default, first, Night and Paper), Slate (the same system in grey-blue with an azure accent, the default of 0.1.0-beta.14 and beta.15), Graphite and gold and Studio ink; the themes of earlier versions, redrawn in the new look — Cantus Night · Paper (espresso and the logo's copper, verdigris on the selection), Blue Hour · Daylight (deep navy, and white with blue), Pewter · Chalk (slate grey with orange and lime, and its chalk twin: the old Slate and Paper, renamed), Ink & Gold · Parchment, Ultraviolet and Lagoon; the new ones — Evergreen Night · Paper (forest green and brass, with real patch cords in the mixer), Bordeaux (wine and rose gold) and Print (black ink on white paper and one vermilion); and the high-contrast pair (every text at 7:1). A click sets a family in both lights; a theme made in one light (Ultraviolet, Lagoon, Bordeaux, Print) also switches the window to that mode. Then the mode (Night, Paper, or follow the system), the cable style, the taskbar icon's colours, and the recolouring (hue, saturation, brightness, contrast, every colour role, or click an element to recolour it). A theme you chose is kept: a setting saved by an earlier version opens in the theme it names, with your own colour changes kept, and themes you made yourself are untouched. If you never chose one — Slate untouched in a setting of beta.14 or beta.15 — Cantus opens in today's default, Slate Lapis; one click on Slate's picture brings Slate back, and it stays. In the high-contrast themes and your own older ones the faders still offer two caps (Faders: Classic and Metal); every other theme draws one cap, lit from above.

Layout and controls (Settings ▸ Appearance, under the look): five choices of this computer, not of the song, each starting where most people want it —

Instrument pictures (Settings ▸ Appearance ▸ Instrument pictures): how Chorda's piano and Tympanum's kit are drawn in their rooms. Drawn (the default) is light and fast: one picture made when something changes, nothing drawn at rest. Photographic adds traced reflections (the lid mirroring the frame and the strings, the floor mirroring the instrument) and soft shadows; it takes a moment to draw — about a second for a full picture, on a background thread that never competes with the audio — and shows a quick version while you drag a control. The sound never changes; the choice is this computer's, not the song's.

The size follows your screen until you choose one: about 125 % on a 1920-pixel screen (less when Windows already scales), never under 100 % nor over 200 %. Ctrl+Plus, Ctrl+Minus and Ctrl+0, or Settings ▸ Appearance, set your own, which is kept; a size you chose in an earlier version is kept too.

Typing a name or a value works the same everywhere — renaming a track, a clip, a pad or a strip, a search, a level: the field opens with its words selected, so typing replaces them; Enter or a click elsewhere keeps what you typed; Esc puts back what was there. While you type, no shortcut fires. Names can be in any script — Japanese, Chinese, Korean, Hindi, Thai, Arabic, Hebrew…: Cantus uses your system's own fonts for them (see the known issues about right-to-left scripts).

Demo songs. File ▸ Demo songs… (Ctrl+Shift+N) opens seven songs made entirely with Cantus's own instruments and effects: With All Your Heart (cinematic pop — the modelled grand Chorda pedalled from its automation lane, every modulation and dynamics effect of round 28, Tonus hard-tuning the topline, modulators and automation, mastered to −11 LUFS), Neon Tide (house), Morning Light (pop ballad), Sunday Strut (funk and soul), Paper Lanterns (lo-fi hip-hop), Aurora (cinematic) and Midnight Static (trap). Open one, play it, take it apart.

The command palette

Ctrl+Shift+P (View ▸ Command palette…) opens a field near the top of the window that finds anything Cantus can do. Type a few letters — of any word, in your language or in English: mix finds Show the mixer, new tr New track with…, spl Split clip — and the list narrows as you type; the arrows move, Enter runs the row, Esc closes. It holds:

What you choose often and lately comes first; with nothing typed, the list starts with what you used last.

The undo history

Alt+H (View ▸ Show the undo history) shows every step of the project's history, beside the inspector — dock it, stack it or float it like any area. Each step is named by what it did — Move 3 notes, Insert Concentus on Lead, Rename track to Bass — under a first row for the project as it was opened. The current step is lit; the steps after it, the ones Redo would bring back, are dimmed; the step your file holds says SAVED, and the steps since the last save carry a dot. Click any row to go there: Cantus undoes (or redoes) each step between, as Ctrl+Z would — nothing is lost, and a click on a later row comes back.

The top bar

The top bar is one row, 44 points high. Every key sits in a recessed well on its centre line; a key that is on is tinted inside its well — green for play, red for record, the accent for the rest. The groups pack in two clusters: at the left the menus, Pat / Song, the transport, the metronome, the tempo, the clock, the grid, undo and the windows; at the right the system meters, the inputs, the master meter, the audio device, the tools, Settings (the gear) and More. At 1920 × 1080 and 100 % only the tools wait in More (in Spanish the inputs too); at a larger size more groups fold.

More. On a narrower window (or at a larger size) the groups fold in a fixed order — the tools, the inputs, CPU and memory, the windows, then compact forms of the device, the metronome, the meter, the tempo, the clock and the menus — and wait in More, which shows how many are folded. Its panel lays them out in sections: Metronome (its key, switch and shortcut; count-in, level, sound, accent), Windows (each key with its shortcut under it), Performance (CPU, memory, voices and dropouts as graphs of the last 30 s with their numbers), Inputs (the ports, the armed tracks by name) and Tools and zoom (− 100 % +); smaller pieces under Also here. Everything works there as on the bar; a group comes back to the bar as soon as it fits. The transport always stays. Customise top bar… (at the foot of More, in View, or right-click the bar): tick a group to show or hide it, drag it by its grip to move it, Reset for the default. Your order is kept.

4. Record audio and MIDI

An audio track

  1. Add an audio track: Ctrl+Shift+A, or right-click the + above the track headers ▸ Add audio track. (The first-run setup's Record on … makes one for you.)
  2. Arm it: click the round red lamp on its header (or Alt+R). Shift+click arms that track alone. An armed track shows its input meter.
  3. Choose its input: the input chip on the header (In 1, In 1-2…; shown when the track row is tall enough), the inspector's Recording section, or the track's menu.
  4. Monitoring: the speaker key on the header cycles Off, Auto (you hear the input while the track is armed — the default) and On. The inspector also has Trim (−24 to +24 dB before recording).
  5. Press R (or the record key). Every armed track records; press R or Space to stop. The whole pass is one undo step.

Takes are written as 32-bit float WAV files: into the project's Audio folder once it is saved, until then into %LOCALAPPDATA%\Cantus\recordings\; saving moves them into the project. Each take is placed earlier by the measured latency, so it lines up with what you heard.

Before FX or after FX (round 40). The inspector's Recording section has Take from: Before FX · After FX. Before FX (the default) records the input as it arrives, after its trim — the strip's inserts change what you hear, not the take. After FX records the strip after its inserts (before its EQ, pan and fader), in stereo, as you hear it — Logic's and Studio One's "record with effects". The take still lands where you played it: a latent insert's delay is taken out. Two things to know: the inserts hear the input only while it is monitored (Auto or On), and they work on everything the strip carries at that moment. The choice is saved with the song, per track; one undo step.

Recording options

The ▾ beside the record key (or right-click it) opens Recording options:

Recording options

Monitoring with low latency

Delay compensation keeps every track in time with the slowest one: with a linear-phase EQ somewhere in the song, everything — the input you monitor included — waits for it, and a singer can hear themself 96 ms late. Low-latency monitoring (Alt+Shift+R, or its switch in the record options, in the audio chip's popover and in the inspector's recording rows) changes that while it is on:

For the lowest delay, monitor through a strip of your own: a bus that mixes your voice with other recorded tracks keeps those in time, and your voice waits there with them.

Takes and comping

Every take is kept on a take lane under the track: open them with the small disclosure at the bottom of the track's colour cell, Alt+T, or the track menu's Show takes. Drag across a take to comp that stretch into the track; click a take to use all of it. Right-click a lane for Use this take, Delete unused takes (which also drops recordings nothing plays any more), Explode takes to new tracks and Hide takes.

Seams. Where two takes meet in the comp they crossfade: equal-power, 10 ms by default, centred on the seam, each take reading on into its own recording. A seam has a small diamond at the foot of the track; drag it (or the edge of either take there) and both sides move together. Right-click it for its crossfade — Off, 5, 10, 20 or 50 ms — or to use that one for every seam of the track. Each piece of the comp shows the number of its take, and the take lanes light what is in the comp.

Alternate comps. The first take lane's header has the comp's chip, Comp A. Click it for the comps: switch between them (each comes back exactly as you left it, pitch edits included), New comp (a copy of this one, to try another idea), Rename, Delete, Flatten (keep the comp, drop the takes) and Flatten and merge (the comp bounced into one recording beside the takes — their files stay on disk). The same menu turns Quick swipe off (then dragging on a lane selects time and a click selects a take) and grouped comping on or off: tracks whose strips are in one mixer group — a drum kit's microphones recorded together — comp together, the same stretch of the same take on each; their seams move together, take the same crossfades, and their comps switch and are made together. Every change is one undo step.

What belongs to the comp. The pieces a swipe or a promote makes are the comp's, and so is the take that plays over the others after a recording (the last pass of a loop, or a take recorded over older ones; a recording where no take lies is a plain clip that plays under every comp): they switch with the comp, and stay its own when split, nudged or slipped. A copy is never the comp's, wherever it is pasted — right over its take too: it stays on the playing lane through every switch, under every comp. A piece dragged to another track leaves the comp; after a flatten the pieces are plain clips. The take lanes light the stretch of each take the comp plays.

MIDI

Arm an instrument track (one with no audio input) and press R: notes from every MIDI keyboard Cantus found (the top bar's MIDI lamp lights as notes come in; its tooltip names the inputs) are recorded into a clip. With nothing armed, R records into the selected track. No keyboard? Ctrl+T turns the typing keyboard on: A–L play notes, Z/X change octave. While it is on, letter shortcuts (including R) are off — use Space or the record key. The keyboard's pitch bend, mod wheel, pedals, every other controller (CC 0–119), channel pressure and how fast you let each key go (release velocity) are recorded too: they land in the clip's lanes under the notes in the piano roll.

MPE: per-note expression

An MPE controller (a Seaboard, a LinnStrument, an Osmose, a Sensel) gives each note its own pitch bend, pressure and timbre (slide, CC 74). Cantus records them on the notes, shows and edits them in the piano roll, and plays them to its instruments and to plug-ins.

MIDI controllers

Any knob, fader, button or pad of a MIDI controller can drive anything Cantus can set: a parameter of any device (Cantus's own or a hosted plug-in's), a fader, a pan, a width, a send, a modulator's amount, a mute, a solo, a record arm, an insert's bypass or mix, the tempo, and the transport (play, stop, record, loop, the metronome, tap tempo, the markers, PAT / SONG — anything a key can do).

Learn in two clicks. Right-click a control ▸ Learn MIDI controller, then move a control on your hardware: done. Or click the top bar's MIDI cell (or press Ctrl+Shift+M; a right-click on it opens the MIDI settings) for the learn mode: every control that can be learnt is outlined; click one, move a hardware control, click the next. A mapped control wears a small dot at its lower right; in the learn mode it is outlined in the highlight's colour, and the one waiting for your hardware is lit. Esc lets go of a learn (so do 15 seconds without a move). With no MIDI input open, a click on the cell opens the MIDI settings instead. Change MIDI controller and Forget MIDI controller are in the same menu. The mixer's strip and insert menus and the track header's menu have Learn rows for the volume, pan, width, mute, solo, arm, bypass and mix, and a strip's for its effects switch and record arm (the strip's FX bulb and record dot learn in the learn mode too); a send's menu has them for its level; the channel rack's knobs and lamps, and a device's switches and choices, learn like any knob. In the inspector an insert's lamp (bypass) and ring (mix), the inserts' switch and the arm LED learn too, and so does a device window's bypass lamp. A control that stands for one value of a choice learns that value: a band's Solo LED learns that band (a button then solos it and lets it go), a key of Tonus's scale learns its click. Inside a plug-in's own window, move the parameter there, then press Alt+Shift+A (Link the last tweaked parameter to a controller, FL Studio's way — also in the top bar's Clip menu) and move the hardware.

Cantus tells what you moved: a plain knob, a 14-bit fader (its fine controller came too), an endless encoder (and which of the three encodings it speaks), a button, a pad (on a switch it toggles; on a knob its velocity sets the value) or a program change (the bank select a keyboard sends before it is passed over). Press a button and let it go: pressed and released it toggles; a button that stays down by itself (it sends its top value only) sets the target on and off as it is. Any parameter with two states (a device's on/off, a lamp) is a switch: a button sets it at once, with no takeover. NRPN and RPN controls are not learnt — Cantus says so and keeps waiting for another control; set the controller to send plain CCs.

How a mapping reads its control — in the Controllers pane (View menu, or Ctrl+Shift+K), one row per mapping:

A lit dot at the left of a row shows a message just came. Shared marks one control on several targets (allowed: one knob can open two filters at once). A row whose target is gone (a deleted track) is kept, marked Missing, and works again if undo brings the target back. A plug-in's parameter is learnt with that plug-in: after the instrument or the insert is swapped for another — a plug-in, or one of Cantus's own — the row says Other plug-in, still named after the plug-in and the parameter it was learnt on, its range in percent, and its knob is free (learn the new one's parameter; an undo of the swap brings the old row back to life). An insert swapped for one of Cantus's own keeps its bypass and mix, and a parameter the new effect also has. Not loaded says the plug-in is in the project but not running here (or it stopped); Frozen, that the target's sound is frozen — a move is refused and said, and the knob reaches no other instrument; Not connected, that the mapping's MIDI input is not open; No input, that a strip's record arm has no track with an input to arm.

Two maps. This project's mappings are saved with the project, and every change is one undo step. Everywhere — your global mappings, kept beside your keyboard shortcuts — holds what means the same in every song: the transport, the tempo, the master, the selected track and the selected channel (volume, pan, mute, solo, arm — the master's when the master is selected) and the focused device's first eight knobs (the device window you clicked last). Learning a transport key or the master puts it there; Add a mapping in the pane offers all of them, and every keyboard shortcut's action by its group (arm the selected track, punch, undo, save, the views…). When one control is in both, the project's mapping wins. If the global file cannot be read whole, a copy is kept beside it (the status says its name) and nothing is lost; if not even a copy can be made, it is never written over, and each change to the global mappings says it will not be kept.

While you play and record. A mapped knob no longer reaches the instrument (it does not also bend its filter), and a mapped pad plays no note. A move made with the mouse or a controller is one undo step per gesture. With a parameter under an automation clip, the clip plays it — until a recording pass moves it: from that moment the parameter follows your hand (the clip is set aside for the rest of the pass), and the pass writes what you did as an automation clip, each move at the moment its message arrived, inside the pass (the count-in and anything past the punch-out are not written, and past the punch-out the parameter follows its clip again; Record options ▸ Record automation). While audio records, a knob mapped to the tempo does nothing (the status says why): change the tempo before or after the take.

Channel pressure and key pressure now reach the instrument (Cantus's and hosted ones), and a program change reaches a hosted instrument.

Cantus follows one outside clock at a time, and sends its own clock to the ports you choose. The top bar's Sync group shows it: LINK (its lamp lit while Link is on, the number of other apps in the session beside it), Q (Link's quantum, in beats) and EXT (the source followed: its lamp is the lock light — lit when locked, amber while pulling in or waiting for its moment — and the tempo it plays). Customise top bar… shows or hides the group like any other.

Ableton Link. Click LINK: Cantus joins the Link session of every app on your network (Live, Bitwig, Traktor, phone apps, another Cantus). The session's tempo becomes Cantus's while Link is on — the project's own tempo and its tempo track are kept and come back when Link is off (as in Live, Link overrides them while it is on). The top bar's tempo shows the session's; dragging it changes the session's tempo for everyone, and so does a change of the project's tempo. Press play and Cantus starts on the next quantum (Q: a click steps 1, 2, 3, 4, 8 beats; 4 is a bar of 4/4), in phase with the others; nothing already played moves, and the sound leaves the speakers when the session's beat does (the audio device's latency is allowed for). With Start/stop sync on (Settings ▸ MIDI), a play or a stop in any app starts or stops all of them. The first time Link is turned on, Windows may ask to let Cantus through its firewall (Link talks to the other apps over UDP on your network, as Live does): allow it on private networks. If Link cannot open the network, its lamp turns red and its tooltip says so.

MIDI clock in. In Settings ▸ MIDI, tick Clock on the input the master device is on (or choose Follow ▸ Clock and the port). Cantus then waits for the master: its Start plays from the beginning, Continue from its song position, Stop stops; the tempo follows the master's clocks, filtered so a jittery clock still plays steadily. Your own play key waits for the master too.

MIDI time code in. Tick MTC on the input instead: Cantus chases the code — a full frame locates, quarter frames run it, a short dropout is ridden through, and it stops when the code stops. Start code says which time code is the song's start (an hour, 01:00:00:00, is common); the project's tempo map turns the time into bars.

Clock and time code out. In the Outputs table, tick Clock on a port to send MIDI clock (24 per beat, the song position pointer, start, stop and continue, following the tempo track) and MTC to send time code at the Code rate chosen (24, 25, 29.97 drop-frame or 30) from the Start code. Each output has an Offset (‹ › in half milliseconds, a click sets zero) for a device that answers late; Feedback says whether controller lights and motor faders may use the port.

The ports' table. Each input's Notes and Controls can be turned off (a keyboard that should only send clock, a controller that should only drive the map); keys and pedals already down are still let go. On an MPE port, Controls carry each note's bend, pressure and timbre, so turned off they go silent.

The session view while following. A clip or scene launched from stop waits for the source: with Link, the next quantum (or the session's start, with Start/stop sync); with MIDI clock or time code, the master's Start. A scene's own tempo is proposed to the Link session at its launch, so the other apps change with it.

5. Edit: clips, piano roll, automation

Quick gestures

As in FL Studio, the arrangement opens things with no hunting through menus:

Clips in the arrangement

Scroll with the wheel, Shift+wheel for time, Ctrl+wheel to zoom. Under the lanes, the song map is the horizontal scroll bar: every track a line where it has clips, the view a frame — drag the frame, drag its ends to zoom, press elsewhere to go there.

The arrangement's tools, Follow and the grid

The arrangement's toolbar holds its name, the tools, the grid, Follow and the panel menu …. The tools are a radio set — the chosen one is lit — and the keys are FL's letters and the number row while the arrangement has the focus (the piano roll's tools otherwise):

ToolKeyWhat it does
SelectE or 3Click a clip to pick it, drag it to move it, drag its edge to resize it; a drag on empty lanes selects. In hand at start
DrawP or 1Click an empty lane to add a bar-long clip in the grid cell you clicked; drag to draw it to the length dragged. On a clip it moves and resizes, as Select
SliceC or 6Click a clip to cut it in two on the grid; the cut is drawn under the pointer first
EraseD or 4Click a clip, or drag across clips, to delete them (one undo step)
MuteT or 5Click a clip, or drag across clips, to mute or unmute them
Zoom8Click to zoom in at the pointer, Alt+click to zoom out, drag across bars to fit them

B (or 2) is still the paint tool of the clip browser (see above), and Z zooms to the selection. The other actions of the old row — split, glue, duplicate, mute, repeat, make unique, delete, loop and zoom to the selection, Snap to the grid — are in the panel menu … and in the clip's menu, with their keys.

What the clips show. A clip's content is drawn at four zooms: far, a MIDI clip's pitch contour, a drum clip's pulse and an audio clip's loudness; closer, the notes in rows ranked by pitch (a melody in pitch order), each drum a row, the waveform's peaks and loudness; close, the notes at their length and brightness by velocity, the stereo sides apart (L, R), an automation lane's values beside its points ("−15.0 dB"). The selected track's clips are lit a step; while a track is soloed, the others' clips are dimmed; a clip under the playhead brightens while it plays; a muted clip is hatched. Moving a clip leaves its old place dashed and the clip riding over a soft shadow, with "Drop → 3.1" on the help line; resizing lights the edge and the help line says the length it lands on (Alt stretches, Ctrl ignores the grid). Point at a MIDI clip and a small notch shows in its top right corner: drag it to repeat the clip at its length (one undo step).

Sections, folders, track headers and a track's icon

The tempo track and markers

A song can change tempo and metre as it goes, and carry named markers (round 29).

The piano roll

Select a MIDI clip and show the piano roll (F7, or its tab under the arrangement): it edits that clip's notes.

The piano roll

ToolKeyWhat it does
DrawP or 1Click adds a note, drag moves, drag the right edge resizes, right-click deletes
PaintB or 2Drag across the grid to paint notes
SelectE or 3Drag to select; drag notes to move them
DeleteD or 4Click or drag over notes to delete
MuteT or 5Click a note to mute it (it stays, silent)
SliceC or 6Drag a line to split notes
PlayY or 7Click notes or rows to hear them

Up/Down transpose the selection (Shift: an octave), Left/Right nudge it, Q quantizes, Ctrl+C/Ctrl+V copy and paste. Velocities are stems in the lane under the grid: drag them, or Alt+wheel over a note. A drum kit shows one named row per pad.

The keyboard and the notes. The keyboard at the left is drawn as a piano: ivory and ebony keys at a real piano's widths (C to E share five rows, F to B seven, so only E|F and B|C fall on row lines), every C named in its key, middle C heavier. The key under the pointer is lit, the key a click sounds is lit more, and while the song plays the keys sounding light in the track's colour. The corner above it, Rows, sets the rows' height: Compact 8, Default 12 or Tall 18 (Ctrl+Alt+wheel over the notes zooms them freely). A note's fill follows its velocity, with a brighter attack and its name inside once it is wide enough; a selected note is lifted and ringed; a muted note is hollow.

Ghosts (the Ghosts lamp in the head, on by default) shows the other clips' notes faintly behind this one — the same track's in its colour, other tracks' in grey — and the head names whose they are ("Pluck · Lead · Choir"). They are never selected or moved.

The head says which clip is open: the track's stripe and name › the clip's name, a menu of this track's other clips and the other tracks' (Other tracks), and, with notes selected, how many and their range ("6 notes selected · E4 – C5 · velocity 90 – 95"). Its tools, grid, Snap, Ghosts and the scale's parts fold on a narrow window, as the arrangement's do.

Lanes under the notes. Under the grid, each clip shows its velocity lane and the lanes it holds. The + in the velocity lane's cell adds one: Release velocity (how fast each note is let go — drag its stems; a ring on the floor is a note with none), Pitch bend, Channel pressure, the named controllers (mod wheel, breath, expression, sustain, …) or Other controller ▸ for any number from 0 to 119. In a lane:

In a laneWhat it does
Pencil, line, curve (the keys in its cell)Draw freehand; drag a straight line; drag a curve (Alt bends it the other way). Line and curve snap to the grid
Drag a pointMoves it (and the points selected with it); Ctrl keeps it in time
Shift + dragSelects the points in the rectangle; Delete deletes them
Right-click a pointDeletes it
× in its cellRemoves the lane (Undo brings it back)

The bend lane shows the instrument's bend range at its edges. A lane belongs to the clip: it moves, splits, loops, copies and links with it; the bend, pressure and pedals go back to rest where the clip ends. With Lanes follow the notes (on, in the + menu), moving, copying and pasting notes takes the controllers under them along. The rows share the space under the splitter; when there are more than fit, the wheel over their cells scrolls them. MIDI files bring their bends, pressure and controllers into these lanes and take them out again on export, with each note's release velocity.

The channel rack

The channel rack (F6) shows every track as a row of steps over the selected clip's bars: click a step on, right-click it off, drag to paint. A step is an ordinary note.

The pitch editor

Double-click an audio clip — or its menu ▸ Edit pitch, or Edit pitch in the inspector — and its take opens in the piano roll's place, Melodyne's way: each note is a blob on the row it sounds, as thick as the voice is loud, with the pitch curve running through it (what was sung, dim, under what will sound). Sibilants and breaths are marked along the bottom. The take is analysed beside the window the first time (a second or so per minute of audio; a recorded take is analysed as soon as it lands) and the result is kept in Cantus's cache, so it opens at once next time. The file is never changed: the clip plays a render of your edits, made beside the window, and while a render is on its way you hear the take as it was (never silence). Export waits for every render.

The pitch lane: right-click an audio track's header ▸ Pitch lane shows its notes in a lane under it, in the arrangement — drag one up or down to move it, double-click it to pull it onto its note — for small fixes without opening the editor.

Formants are kept when a note moves (measured within 3 dB at ±3 semitones). It works on one voice or one instrument line at a time; chords are not split into notes.

Warping and stretching audio

Warping makes an audio clip play in the project's time: faster or slower than it was recorded, without changing its pitch, and following every tempo change of the tempo track. The file is never changed: the clip plays a render made beside the window.

While a render is on its way (a fraction of a second for a loop, a few seconds for a long file) you hear a quick preview at the right timing and pitch — a clip transposed for the first time plays its own audio meanwhile; one whose markers move its audio waits a moment for its preview, never playing at the wrong bar — and the change from one to the next is crossfaded once. Export, stems, freezing, bouncing and a modulator's bake wait for the full render; if one ever takes too long, the status line says so when the job is done. Warped clips travel in DAWproject files with their markers and mode (a file Cantus wrote comes back with its warp exactly as it was); a clip at its natural speed travels as a file placed in seconds, so it keeps its speed in any application. Opening or dropping a .dawproject reads it in the background (its audio copied to disk as it goes, a file referenced beside it found there); it opens once nothing is being recorded or dragged, and never over another project opened meanwhile.

Automation

Right-click any knob, fader or effect control ▸ Create automation clip, or move a control and press Alt+A (automate the last thing you touched). The clip covers the time selection (else the loop, else four bars) on a lane under its track. On the lane: drag points, drag the ring between two points to bend the curve, click the line to add a point, double-click a point to delete it, right-click it for its shape (single or double curve, hold, stairs, pulse, wave). Double-click the clip for the large editor, with a drawing tool and articulators (flip, smooth, random, scale, LFO). An automated control shows a dot.

Hosted plug-ins are automated like Cantus's own: right-click a parameter in a plug-in's generic editor (or in a hosted instrument's Parameters list in the inspector), or move it in the plug-in's own window and press Alt+A. Their changes reach the plug-in at their exact sample inside the audio block.

Recording automation. While you record, every control you move — in Cantus or in a plug-in's own window — is written into an automation clip over the stretch you moved it in (one undo step when the pass ends). Switch it off in the record options (Automation). (This is the Latch mode below, for every track in Read.)

Writing automation while the song plays: Read, Write, Touch, Latch. Each track has an automation mode — the badge at the right of its header (R, W, T, L; click it, or Alt+Shift+W for the selected track), the inspector's Automation row; the master's is in its menu on the mixer. The automation arm in the top bar (the envelope key after loop) makes them write; its ▾ sets every track at once and the Touch glide.

ModeWhile the song plays, armed
Read (R)Plays the automation. Moving a control writes nothing
Write (W)Overwrites everything the track automates, from play to stop, each held where it was until you move it. After the pass the track switches to Touch
Touch (T)Writes while you hold a control; let go and it glides back to the curve that was there (over the touch glide, 250 ms by default)
Latch (L)Writes from your first touch until you stop, holding the last value

What you write becomes ordinary automation clips over exactly the stretch you wrote; the control goes back to where it was before the first clip, and the whole pass — with the moves of your hand — is one undo step. It works for every automatable control: faders, pans, sends, effects' and instruments' parameters, plug-ins' parameters moved in their own windows, modulator depths, and controllers mapped to any of them. Round a loop, each pass writes over the last where it touches; with punch on, only the punch range is written.

Modulators

A modulator moves parameters for you: an LFO (five shapes, synced to the song or free in Hz, restarted by notes if you like), an envelope (attack, decay, sustain, release — opened by the track's notes or by an audio level), a follower (the level of any strip: duck a pad under a kick), steps (up to 32, synced, with glide) and a random source (held or smooth, repeatable from its seed). Each modulator drives any number of parameters, each with its own depth, plus or minus; a parameter may take several, and they add up on top of its automation. Two more are controllers you play or program, each with its own window: Tactus, the X-Y pad, and Calculus, the formula controller.

Synced modulators follow the song position, so playing from bar 5 hears what a full play hears there; free ones start with playback. Exports sound exactly as playback.

Tactus, the X-Y pad

Tactus (Latin for touch) puts two parameters under one finger: drag the puck across the pad and the horizontal position (X) moves one, the vertical (Y) another — a filter's cutoff and resonance, a pan and a width, two sends — like FL Studio's X-Y Controller or a Kaoss pad. It is a modulator, so it lives where the others do and drives anything they can, as many parameters as you like, each through the axis and the range you give it.

Make one: in the inspector's Modulators section click the band's + ▸ X-Y pad, or right-click a knob ▸ Modulate… ▸ X-Y pad (it comes already connected to that knob). Open it: the Open key on its row (Window), or double-click its picture. The window floats like an effect's editor: drag it by its header, close it with its key or Esc, minimise or maximise it, scale it from its corner grip. The lamp in the header switches the pad on and off, and its ⋯ menu offers three colourways — Titanium, Copper and Frost (the formula's: Bakelite, Ivory, Ebony). One controller window is open at a time; it closes by itself when its modulator is removed.

Connect what it drives as any modulator: right-click a parameter ▸ Modulate… ▸ the pad, or drag the pad's handle (the grip at the left of its title in the inspector) onto a knob. Then set each link in the window's What it drives list, one row a parameter:

Each drag in the list is one undo step; so is an axis click or a reset. The list grows with its rows and scrolls with the wheel when there are many; the pad's rails name the first two parameters each axis drives and count the rest.

Play it. Drag on the glass: the puck follows, each axis moves its parameters, and the whole drag is one undo step. The ring is where the puck is now; the crosshair is where it heads (your hand, the automation, or the path); a fading trail follows it.

Record a gesture as automation. While a pass writes automation — recording with the record options' Automation on, or the automation arm with the track in Write, Touch or Latch (see Automation) — drag the puck: Cantus writes two lanes, X and Y, over exactly the stretch you moved it in, and the pass is one undo step. Played back, the lanes move the puck as your hand did; edit them like any automation. To automate one axis by hand, right-click its rail (the words along the pad's foot for X, along its left side for Y) or its knob on the Pad card: the menu every parameter has — Create automation clip (Edit automation once there is one), Modulate…, Learn MIDI controller (Forget once linked), Reset to default. Or move its knob and press Alt+A (automate the last thing you touched); Alt+Shift+A links it to a MIDI controller the same way. Amount offers the same; the inertia, the path's settings and the polarity are the pad's own settings, not parameters, and offer none of it.

The last thing touched is the pad. After a drag on the pad, Alt+A makes the lanes of both axes, X and Y, in one undo step (as a recorded gesture writes two); when both already have theirs, it shows them. Alt+Shift+A links both to your MIDI controller: move a control for X, then another for Y (the status line says which it waits for; the control just learnt for X is passed over while Y waits; Esc lets both go). Each link is an undo step. A knob on the Pad card is one axis: Alt+A after it makes that axis's lane alone.

Paths: gestures that loop with the song. A path is a loop the puck follows while the song plays (stopped, your hand has the pad again), read from the song position — so a play from bar 9 hears what a full play hears there, and an export sounds the same.

Drawing or recording a path is one undo step, the moves of a recording with it.

Presets — the well in the header: its arrows step through them; click the name for the list, each with what it does. Choosing one keeps the pad's connections and sets the axis and range of its first routes in order — so connect the parameter you want on X first and the one on Y second. The name stays while you play the pad and reads Custom once you change a setting.

PresetWhat it does
Filter sweepLink X to a filter's cutoff and Y to its resonance: X opens it all the way (one way, left the knob's value), Y adds up to 60 %, bent; an 80 ms glide
Width and panLink X to a pan and Y to a width: from the centre, X pans both ways and Y narrows or widens about the knobs; 120 ms
Figure eightA smooth figure-eight path over two bars that plays with the song: link X and Y to any two parameters
Performance padBoth ways from the centre with a slow 400 ms glide: for the two parameters you play live

Bake to automation works for the pad as for any modulator (a route's bake key, or Modulate… ▸ Bake to automation): the clip is what the route played — the axis, the range, the path, the inertia and any automation of X and Y included. AI agents make, connect and play pads too (What an AI agent can do). With its puck in the centre (Both ways), or connected at depth 0, a pad changes no sample.

Calculus, the formula controller

Calculus (Latin for a reckoning stone) moves parameters by a formula you type — of the song's position, its tempo and three knobs a, b and c — in the syntax of FL Studio's Formula Controller: 0.5*Sin(SongTime*Pi)+0.5 is a sine over two beats. One formula drives any number of parameters, each through its own range, in time with the song; a render from any bar hears what the full song hears there.

Make one, open it, connect it as the X-Y pad above: the band's + ▸ Formula (or Modulate… ▸ Formula), its Open key or a double-click on its picture, then Modulate… or its handle onto each parameter. In What it drives each row's bar sets that parameter's range and bend exactly as on the pad (a formula has one output, so no X/Y tag).

The window is a scientific calculator:

The value is kept to 0…1 as in FL (anything that is not a number is 0), then the route's depth and range apply.

What you can write. Upper or lower case alike; a name may also be written with empty parentheses (SongTime()).

KindWhat you write
Valuesa, b, c (the knobs, 0…1); SongTime (also beat): the song position in quarter notes; t (also seconds): seconds since playback started; Time: milliseconds since playback started; GetTempo (also tempo): the tempo in BPM; Pi; e; Rand: a random 0…1 for each song position; numbers, with exponents (1e3)
Arithmetic+ - * /, ^ (a power; 2^3^2 is 2⁹, 2^-1 a half), % after a number (a hundredth: 50% is 0.5), ! (factorial), div and mod (integer division and remainder of the whole parts), fmod, shl and shr (times or divided by a power of two)
Comparisons and logic= <> < <= > >= and in (1 when true, else 0); and, or, xor, not (above 0.5 counts as true)
FunctionsSin Cos Tan (Tg) Sec CoSec Ctg (Cot) ArcSin ArcCos ArcTg (ArcTan) ATan2(y, x) Sinh Cosh Tanh ArcCosh ArcTanh DegToRad RadToDeg; Int (Trunc, toward zero) Floor Ceil Round (halves to even) Frac (the fraction, with its sign) Abs Neg; Sqrt Sqr (the square) Exp Ln Log10 Log2 LogN(base, x) Power(x, y) (Pow) IntPower(x, n); Max(x, y) Min(x, y) Sum(x, y); Fold (reflects any value into 0…1)
ChoicesIf(test, then, else); Case(a, b, c) (b when a is 1, else c); IfE IfG IfGE IfL IfLE(x, y) (1 when x =, >, ≥, <, ≤ y); Inter(x, lo, hi) (1 when x is between them; InterNoSwap when lo ≤ x ≤ hi as written); Map(x, lo, hi) (x from 0…1 to lo…hi); clamp(x, lo, hi)
RandomRand (above); SeededRand(n): a random 0…1 for each whole number n — the same every time
Cantus's waves0…1, one cycle for each whole number of their argument: sine(x) (starting at 0), saw(x), tri(x), square(x) (1 for the first half), pulse(x, width) (1 for the first width of the cycle)

Three of FL's names behave differently, so that every render sounds the same: Time counts from when playback started (FL's reads the computer's clock), Rand comes from the song position (FL's is a running generator) and SeededRand is Cantus's own sequence. Three are refused because a render would depend on the moment it was made — MouseX, MouseY, Date — and two because FL's manual does not say what they compute: Tension and Warp. A formula is one expression, up to 512 characters; one too complex is refused with shorten it or nest it less.

A few formulas (for a 4/4 song):

FormulaWhat it draws
0.5*Sin(SongTime*Pi)+0.5A sine over two beats (the formula a new Calculus starts with)
Frac(SongTime/4)A ramp up each bar, falling back on the bar line
square(SongTime)On for the first half of every beat, off for the second
SongTime mod 4 < 1On for the first beat of every bar
SeededRand(Int(SongTime*2))A new random value every eighth note, the same in every render
Map(sine(SongTime/8), 0.2, 0.8)A slow swing between 20 % and 80 % over two bars
If(SongTime < 32, SongTime/32, 1)A climb over the first eight bars, then held
a*Sin(SongTime*Pi*b)+cFL's knob example: a the size, b the speed, c the centre

Presets (the header's well, as on the pad) share one plan for the knobs: a is the rate (from one cycle in four bars to four a beat, in steps of a doubling: 0.34 is a bar, 0.67 a beat), b the shape and c the depth. Their texts are ordinary formulas — open one and change it.

PresetWhat it does (a · b · c)
Sine, Triangle, Saw, SquareBoth ways about the knob's value, a bar a cycle; a the rate, b the phase, c the depth
PulseA short high, then low, each beat; b the pulse's width
Random each beatA new random value every beat, the same in every render; b another sequence
Gate on beatsOpens on every beat for the width b, with soft edges (one way: for a level or a send)
Envelope each barStruck on every bar line, then dying away; a the attack, b the decay
ClimbRises from the song's start over a (1 to 64 bars), bent by b, to the height c — a build-up

A preset sets the formula, the knobs and the polarity, keeps the connections and lets every route read the whole formula (no range); its name stays while you turn the knobs.

Bake to automation turns a route into the clip it played (knobs, automation and range included). AI agents write formulas too, and one that is not a formula is refused with the character it fails at (What an AI agent can do). A formula connected at depth 0, or with nothing connected, touches no sample.

Note effects: arpeggiator, chords, scale

A track can change the notes it is given before its instrument plays them: an arpeggiator, chords and a scale, in a chain, in the order you set (FL's channel Arp and Chord, Ableton's MIDI effects). They act on the clips' notes and on what you play live, alike — while the song plays and while it is stopped.

Note effects in the inspector

The song's key, and the piano roll's scale

Set the song's key on the piano roll's toolbar: click Key and pick a root and a scale (major, minor, harmonic and melodic minor, the modes, the pentatonics, blues, whole tone, the diminished scales, Phrygian dominant, Hungarian minor, Japanese, chromatic) — or No key. It is saved with the song.

The piano roll in D minor

Grooves

A groove gives a clip a feel — notes late or early, accented, longer or shorter — without touching its notes. In a MIDI clip's inspector, Groove opens the list: the built-in swings (8ths and 16ths, MPC-style, 50–75 %) and the project's own grooves.

A clip's groove

A MIDI file export asks which notes to write: As played (each clip's groove applied and each track's note effects written out note by note — the file plays what you hear), or As written (the clips' notes, no groove, no note effects).

The score: notation, MusicXML, PDF and printing

Every note clip and every MIDI track can be read as an engraved score, made as Dorico, MuseScore and Sibelius make one — the notes read onto values the metre shows, tied where a beat or the half bar must be seen, beamed by the beat, spelled in the song's key with accidentals per bar, in the SMuFL font Bravura.

The session view: launching clips and scenes

The Session pane is a clip launcher, as in Ableton Live and Bitwig: the same tracks as the arrangement, side by side as columns, and scenes as rows. Each cell (a slot) holds a clip you can start and stop while the song plays, in time; a scene starts a whole row at once. The tracks, the mixer and the instruments are the arrangement's — the session only decides which clip each track plays.

The session: how launching, follow actions and recording work

The session view's own page tells you where everything is; this one tells you what happens when you press it.

Same tracks, same mixer. A column of the session is a track of the song: the same instrument, the same mixer strip, the same automation. A clip launched on a track takes over that track's arrangement until you press Back to Arrangement (on the track, or for every track) — even after the clip stops, the track stays with the session, as in Live.

When a launch happens. A clip waits for its launch quantisation — its own, or the session's global one (a bar by default) — and starts exactly on that boundary, to the sample. A launch pressed just before a boundary, too late for the engine to have it ready there (a few milliseconds to tens of them, by your audio buffer), still plays on it — a moment late, caught up to where it would be, never a whole bar later; and so does one pressed a 32nd note or less after a boundary (on a launch grid of a 32nd or finer, a quarter of a grid step). With the transport stopped, a launch starts it at once, from the playhead. A scene launches its whole row on the global quantisation, and sets its tempo and metre there (the metre on the next bar line).

Following Link, MIDI clock or time code. A launch from stop waits for the source to start: Link's next quantum, or the master's Start — however long that is; the clip shows as queued meanwhile, and the stop button calls it off. If the master starts from another place (MIDI clock's Start is from the top), the clip plays from there. Following Link, a scene with a tempo sets the tempo of the whole Link session when it takes effect, as in Live. A relock of Link, or the master moving the playhead, does not end what the session plays.

Launch modes. Trigger starts the clip (again) on each press. Gate plays it while held and stops it at the boundary after you let go. Toggle starts it, and stops it when pressed again. Repeat restarts it on every boundary while held. Legato makes a launched clip take over where the clip it replaces had got to, instead of starting from its start marker.

Follow actions. After its follow time (a time, or linked to its own length, times its count) a clip does action A or B — stop, play again, the previous, next, first, last, any or another clip of its column, or jump to a row — with the chance you set for A. The chances are drawn from the session's seed: the same performance plays the same way every time, and an export renders exactly what you heard. Change the seed for another performance.

Recording into a slot. On an armed track, the record button on an empty slot records from the next boundary (after the count-in, from stop) to the bar line after you stop it — or for the session's fixed length — and the new clip plays on at once, looping. MIDI is quantised by the session's record quantisation. Recorded over a looping MIDI clip that plays, the notes go into that clip (overdub). The take is kept like any recording: a crash loses nothing.

Recording the session into the arrangement. With arrangement record on, everything you launch, stop and every follow action is written into the arrangement exactly where you heard it — the clips, their envelopes as automation, the scenes' tempos — as one undo step, when you switch it off or stop. The arrangement then plays exactly what you heard, to the bit.

Good to know. A frozen track's clips do not launch (unfreeze it first). Freeze, bounce and export render the arrangement: record a performance into it first to keep it. Stopping the transport, or jumping the playhead, ends the performance. A tempo change keeps an unwarped audio clip playing the same stretch of its file; a warped one follows the tempo.

Audio to MIDI. Right-click an audio clip ▸ Convert to MIDI… turns a melody, a chord part or a drum loop into notes on a new track: see Convert audio to MIDI.

6. Mix: the mixer, inserts and sends

The mixer

Each track feeds a strip of the mixer (F9); strips feed the master at the left, or another strip (a bus). A strip reads from top to bottom in the order the signal takes: its head (the number over the name, the track's icon, the arm dot), the inserts (a dot and the unit's kind for each; the first free row a faint +), its own EQ curve, polarity and swap, pan (and the width knob), M | S as one pill, the fader with its meter beside it and its level under it, the sends as small knobs, and the output ("Master" quietly, or "→ Drum bus"). Nothing reads "−∞" at rest; there is no label column. Three widths — Compact, Wide and Large (the unit's name on each insert, the send levels in dB, the EQ bands as dots on the curve, the master's short-term LUFS) — on the desk's header or with Ctrl+wheel; a short desk gives way row by row (the EQ, the polarity row, the inserts to one summary line, the sends…) and then scrolls. When the mixer is docked under a bar with room, its header — the widths, Rows ▾ (which rows the strips show, From the selected strip among them, and Insert panel), the solo chip ("1 soloed · Clear"), Cables, Snapshots and + — sits on the workspace's row of tabs, right of the tabs. Click a strip to select it (its body tinted, its stripe whole), Ctrl+click to add, Shift+click for a run; dragging one fader of a selection moves them all. With the mixer focused, M mutes, S solos, Alt+Left/Alt+Right move the strip. A muted strip goes quiet in ink and its M lights in the theme's accent (Settings ▸ Appearance ▸ Layout and controls ▸ Lit mute button: Grey lights it grey instead); a strip silenced by another's solo shows its M ringed in the solo colour. The returns docked at the right sit at the window's edge, as in FL Studio, and the room between the last strip and them holds the next empty strips, numbered — a click on one makes the strips up to it (Mixer's right dock: After the last strip puts the returns right after your strips instead).

Groups, for FL Studio users. FL has no "group" button: you group by routing. Cantus works the same way, with two kinds:

Bands on the desk. A group of strips is drawn as a band over them: a gap and a seam either side, its name and count at the top. Select strips and right-click ▸ Group ▸ New group from selected (double-click its head to rename it). A click on its chevron folds the band into one strip that names its members and shows their meters side by side (Unfold brings them back — a folded group is never hidden); drag its right edge to take in the next strip or to give its last one back (Esc leaves it as it was). Each change is one undo step. An instrument with its own outputs (Chorda, Tympanum) draws its strips as a family — a solid bar with its glyph, Drums · Tympanum · 12 mics — that folds the same way; Each mic to its own strip makes a sub-group for each family of microphones (Kick, Snare, Toms, Cymbals, Overheads, Room) in the same step, and All to Main takes them away. A strip is in one band at most: a group sits inside one family or takes whole families. A strip fed only by a family's outputs — a drum bus, a parallel compressor — can join its band (right-click ▸ Add to …'s band; it is drawn at the band's end, marked added) and leaves it, the status line saying why, when its input stops coming only from that family. While the desk scrolls, a band's name stays in view.

Solo follows the routing: a soloed strip is heard with what feeds it and what it feeds.

The inspector

The inspector (I) shows the selected track, strip or clip as a signal-flow sheet. For a strip, under its head (the colour, the number, the name, Mixer strip and the ⋯ menu):

Then the sections that fold to one line each, with their state at the right — a click on a title opens or closes it:

Freeze and bounce in place

Freeze turns a heavy track into a recording of itself, so its instrument and effects stop using the processor — and gives it back, exactly, when you unfreeze it.

Bounce in place renders to an audio clip, on a new track right under the original, at the same place:

7. Master: the master chain, Fervor, Latitudo, Culmen

The master has its own chain of up to ten inserts, run on the whole mix before the master fader. Click the master's head to see its chain in the inspector (The whole mix → the inserts → Audio output), or right-click its name ▸ Effects ▸ + Add effect.

Culmen on the master

The demos master in this order, and it is a good start:

  1. Aequalis (EQ): a high-pass under the sub, a little air.
  2. Multiplex or Comprimo: gentle glue.
  3. Latitudo (stereo): the preset Master: tighten the low end or Master: wider top, mono bass keeps the bass in mono and the top wide. Watch its correlation meter: it should stay above zero.
  4. Culmen (maximizer): choose a target — Streaming −14 (−14 LUFS, −1 dBTP), Apple −16, CD −9, Club −7, Broadcast −23 — then play the loudest part of the song and click Learn drive: Culmen listens, sets the drive, listens again and stops on the target, in one undo step. The Loudness card shows momentary, short-term and integrated loudness and the true peak; the verdict says On target or how many LU you are over or under. Culmen works in three bands plus the full band, into a true-peak brickwall at the ceiling; its Character (Clean, Punch, Loud, Bus, Safe) and its ten presets cover most masters.

Fervor (saturation) adds colour: Tape, Tube, Overdrive, Fuzz, Clip, Fold or a drawn Curve, with oversampling, tone, mix and Level match, so that louder never fools you. On the master, the presets Master clipper and Tube glue on the mid are the gentle ones.

The mastering assistant

Right-click the master's name ▸ Mastering assistant…, or open the master chain's + Add effect list: the first row is the assistant. It listens to your mix as it reaches the master and proposes a chain of Cantus's own effects to master it — nothing it does is hidden: what it proposes are ordinary inserts you can open and change.

  1. Choose where the song is going: Platform (‹ › steps through Spotify, Apple Music, YouTube, TIDAL, Amazon Music, SoundCloud, CD, Club and EBU R 128). The line under it says what the platform aims at — Spotify plays every song at −14 LUFS and wants the true peak at or under −1 dBTP — and the bottom of the panel says where that number comes from and when it was checked.
  2. Choose the Balance: a genre's curve (Pop, Rock, Electronic, Hip-hop, Acoustic, Classical), or Load reference… — a finished track you like the sound of (WAV, FLAC, MP3, OGG, AIFF). With a reference the assistant matches its tone and width; the platform still sets the loudness.
  3. Listen to the whole song or only the loop. If the master already has effects, choose whether the proposal replaces them or is added after them.
  4. Click Listen and propose. Cantus renders the mix (a song of three minutes takes as long as an export), measures it and tries the chain on it until the loudness lands on the target. Stop stops it.

The table shows, for the mix, the proposed master and the target: integrated loudness, true peak, loudness range, the loudest three seconds, width (how much side there is against the middle), what folding to mono loses, and the lowest correlation. The chart draws the balance per octave — bars for the mix, a line for the proposal, rings for the target. Under it, each unit of the chain and why it is there: an EQ toward the curve (at most ±3 dB a band; ±6 with a reference), gentle glue when the song's dynamics are wider than the genre keeps, stereo width or a mono bass when the mix folds badly to mono, and a true-peak limiter whose gain lands the platform's loudness under its ceiling.

A/B plays the mix (A) or the proposed master (B) from the same place, the louder one turned down to the other's loudness: you compare tone and dynamics, not level — louder always sounds better, and the A/B does not let it fool you. Nothing in the project changes while you compare.

Apply puts the chain on the master in one undo step (Edit ▸ Undo takes it all back). Then play the song and adjust anything by ear — every knob is yours.

Listening for AI agents

When you allow an assistant such as Claude to work in Cantus (Settings ▸ AI), it can listen too: audition renders a range — or a soloed part, or the mix without the master chain — and answers with the numbers above for the range and each section, the spectrum against a genre's curve, clicks at section changes, and a WAV of what it heard (in Cantus's own cache); analyse_tracks measures every track in the mix; masking_matrix says which tracks fight for the same frequencies, and where; set_mix_targets sets tracks to the loudness the agent asks for, in one undo step; master_assist, apply_master_chain and replace_master_chain (which asks you first: it takes the master's own effects out) are the mastering assistant above. Cantus keeps a copy of the song's state at each marker (or every eight bars) while it renders, so after the agent changes the last chorus it hears only the last chorus again — the same sound, to the sample, as playing the whole song. With a hosted plug-in in the project that copy cannot be made; the range then starts eight seconds early, and the answer says so. The prompt Master for streaming asks an agent to do all of it for a platform you name.

8. Export

File ▸ Export audio… (Ctrl+Shift+E in the default keys) opens the export window: presets on the left, the choices on the right, and a line that says exactly what will be written.

Exports run in the background with a progress bar; Stop leaves no partial file. Your last choices are remembered.

9. Sounds: the browser and sound packs

The browser (Alt+F8, the first window key) docks at the left. It lists your Favourites, the sound packs you installed (Packs) and My folders (click + to add any folder of samples).

The sample browser

The Sound packs window (Sound packs…, or More packs…) shows every pack as a cover drawn from what it holds (drum rings, a waveform, a keyboard, an impulse response…), grouped by category, with its name, what it is, how many files and its licence, and its state: Get · 140 MB (a click fetches it); while it downloads, a thin bar, "42 % of 54 MB" and Cancel; Installed with a check, and Open (it opens in the browser's tree); or Update · 4 MB. Its head counts the packs ("8 free packs · 3 installed · 471 MB on disk") and filters them (All · Installed and each category); Sort at the right orders them by category, name or size. Click a cover for its page beside the grid: details, licence and credits, Install, Update, Remove, Show the folder, and Install from a file… for computers without internet. Where packs are kept › at the foot shows the folder. The packs: Cantus's own, Karoryfer's kits, VCSL and VSCO 2 percussion, a TX81Z for Exemplar, the Salamander Grand Piano (a Yamaha C5, 16 velocity layers), DRSKit, an acoustic drum kit recorded by the DrumGizmo team, and the OpenAIR Impulse Responses for Aula.

Every pack is free. Most are CC0 (no credit needed); the Salamander Grand Piano is CC BY 3.0 and DRSKit CC BY 4.0: their credit is shown on their page in Sound packs… and in Help ▸ Credits, and kept in the pack's folder (DrumGizmo also asks, in its own words quoted on the pack's page, for a credit wherever a song's other credits are given). Either way, the music you make with them is yours. Packs are saved in %LOCALAPPDATA%\Cantus\Content\; every file is checked against its SHA-256.

Pack instruments. An installed pack that brings instruments (the Salamander: with 8 velocity layers, about 120 MB of memory, or all 16, about 185 MB) shows them in the device picker's Keys shelf and on its Sound packs… page (New track with …); each plays in Exemplar. Long samples are streamed from disk: only their first moments stay in memory. The first time, Cantus decodes the samples once into its sample cache (1 GB for the Salamander's 8 layers, 2 GB for 16), beside the window, with Loading samples… n of m on the status bar; later loads take a fraction of a second. If the disk cannot keep up, the status bar counts Disk late: n.

The acoustic drum kit. DRSKit (DrumGizmo team and Jes Eiler of DRSDrums) was recorded with 13 microphones; when you install it (a 2.8 GB download), Cantus mixes them into a stereo kit — every drum at its place in the stereo field, the snare's bottom microphone in polarity, the toms' and cymbals' close microphones turned down on the other drums as a mixing engineer's gates would — so each hit plays as one stereo sample. It takes about two minutes after the download and leaves about 470 MB; while it installs it needs about 7.6 GB free (the download, the unpacked kit and the mix side by side — the pack's page says so, and an install without the room is refused before anything is downloaded). If the mixing is cancelled or fails, the download is kept: the next try starts from it. DRSKit acoustic drums is then on the device picker's Drums shelf, playing in Exemplar on the General MIDI drum keys with its own drums only: kick 36 (35 a more open kick), side stick 37, snare 38 (40 a damped snare), toms 41 · 43–45 · 47–50, hi-hat closed 42, pedal 44 and open 46, crashes 49 and 57, ride 51 and 59, bell 53. The General MIDI keys the kit has no sound for (tambourine, cowbell, the Latin percussion) stay silent, so a General MIDI part never plays a cymbal in their place; DrumGizmo's other articulations sit above General MIDI 2's percussion and clear of the XG and GS kits' extra keys below it (13–26): the chained ride on 88–90, the brushes on 91–100, and the half-open hi-hat and the tips on 101–106. Each drum has its strengths as velocity layers down to 24 dB under its loudest, and the hits of a layer take their turns in order — counted per drum and per layer, so a roll, a tom fill across the drum's keys or an accent on every fourth hat never plays the same sample twice in a row; a closed or pedal hi-hat silences an open one. In the piano roll and the channel rack the kit shows a row per drum, by name, in the interface's language (Exemplar's Drum rows switch, in its Edit view). The microphones' balance is fixed by the mix: faders for each microphone need an instrument with several outputs, a later step.

Round robin in order. Exemplar's Edit view has the round robin's order: In order (each note takes the next turn of the samples it reaches — whatever key or velocity reaches the same ones shares the turns — by its place in the song, so a render from any point chooses what the whole song does there) or By note (drawn from each note — how Exemplars in projects and user presets saved before this version keep playing). An imported instrument plays its round robins in order.

A missing pack. A project that plays a pack this computer does not have opens a card that names the pack and its silent tracks: Install in Sound packs…, Use the installed pack (when the project points at another copy, say from another computer), or Find the folder… (when files of an installed pack were moved). The rest of the project plays as usual.

10. Instruments and effects: the device picker

The device picker shows every instrument and effect — Cantus's own and the CLAP and VST3 plug-ins installed on your computer — as tiles.

The device picker

Open it with the + above the track headers (a new instrument track), by clicking a track's instrument chip (swap the instrument), Swap on any device window, an empty insert slot, or View ▸ Device picker…. Search by name, maker or kind; filter by shelf (Favourites, Recent, a category, or Cantus / CLAP / VST3); arrows move, Enter chooses, Ctrl+D marks a favourite, Esc closes. Installed pack instruments (the Salamander Grand Piano) are under Keys, DRSKit under Drums. Swapping an instrument keeps its notes, clips, mixer strip and effects; swapping an effect keeps its slot, mix and bypass.

InstrumentWhat it is
FirmusThe simple synth every new track starts with
Pulsus IIDrum machine: synthesized drums and sample pads, choke groups
UndaWavetable synth: three oscillators, unison, two filters, drawn LFOs
CalorAnalog polysynth
ModulusFM synth with eight operators; reads DX7 banks
ExemplarSampler: SFZ, Decent Sampler and SoundFont
GravisBass: 808, synth and modelled electric bass
LaminaElectric piano
OrganumTonewheel organ with drawbars and a rotary speaker
ChordaGrand piano played from a physical model: strings, hammers, soundboard, pedals
TympanumAcoustic drummer: modelled kits with movable microphones and recorded multi-mic kits, each mic on its own strip (Tympanum)

The effects are Aequalis (EQ), Comprimo (compressor), Porta (gate and expander), Sibilus (de-esser), Multiplex (multiband dynamics), Terminus (limiter), Culmen (maximizer), Tonsor (clipper), Impetus (transient shaper), Discrimen (transient splitter), Spatium (reverb), Fervor (saturation), Latitudo (stereo), Speculum (analyser), Tonus (pitch correction), Concentus (chorus), Flexus (flanger), Phasis (phaser), Tremor (tremolo and auto-pan), Vibrans (vibrato), Anulus (ring modulator), Translatio (frequency shifter), Fluctus (volume shaper), Cribrum (auto filter), Fragor (distortion pedals), Lapsus (time shaper), Vetus (lo-fi), Libra (utility), Aula (convolution reverb), Delay and Reverb.

Each unit's face

Each of Cantus's effects and instruments wears a face of its own — all but the simple Reverb and Firmus, the synth every new track starts with, which keep the kit's plain look — as hardware does and as the best plug-ins do: its colour and the material of its panel (brushed metal, wood, matte paint, vinyl amp covering, glass…), the look of its display (a glowing screen, an LCD, ink on paper, an oscilloscope's grid), the shape of its knobs and switches (a value ring, a skirted knob, a chicken head, a graduated dial; a toggle, a rocker, a lamp button) and the lettering of its name. You recognise a unit at a glance, even small in a corner of the screen.

A face changes how a unit looks, never how it works: every control keeps its name and its value under it, its right-click menu (Modulate…, Learn MIDI controller, automation), the keyboard and the screen reader; the window still resizes from 75 % to 200 %. Faces follow your theme — every face has a dark and a light version, and words stay readable on it.

Prefer every unit in your theme's plain look? Settings ▸ Appearance ▸ Units wear their own faces ▸ Off. The choice is kept, and applies at once to the windows that are open. Hosted plug-ins shown in Cantus's view keep the plain look; the Cantus effects you use as CLAP or VST3 plug-ins in another program always wear their face.

Each instrument's face

The instruments' faces, one by one (the controls, the right-click menu and the keyboard work the same in every one):

InstrumentIts face
UndaDeep indigo glass; the wavetable glows cyan in 3D, the frame playing now in magenta
CalorA warm brushed panel with orange legends, machined knobs, a scope under the filter
ModulusA matte umber panel; operators, envelopes and spectrum on a green liquid-crystal display; membrane keys whose colour bar lights
ExemplarA grey-violet anodised case; zones and waveform glowing violet on a phosphor screen, the loop's points in teal
GravisBlack vinyl amp covering with a blue pilot light, pointer knobs; the pluck marker in red
LaminaBurgundy lacquer in rose; the tine with its brass tone bar and the pickup behind glass
OrganumWalnut with a magic eye's green; drawbars in white, black and brown on a printed chart, ivory stop tabs that tip down and light, graduated dials
ChordaBlack lacquer (white in a light theme) and gold; the strings over a spruce soundboard, the cast frame's rim, the felt hammers
Pulsus IIA dark case with 808-yellow lamp keys; each hit on a yellow fluorescent display, its envelope in orange

Chorda: the modelled grand piano

Chorda is a grand piano computed as you play — no samples. Every key has its own one, two or three strings, struck by a felt hammer and ringing into a soundboard; the strings of the other keys ring along when the dampers let them. So it answers the way a piano does: the harder you play, the brighter; a held chord rings on, the strings beating gently; with the sustain pedal (controller 64) the whole instrument resonates, and half pedalling works — a pedal partly down damps partly. The soft pedal (una corda, controller 67) shifts the hammers for the next notes: softer and darker.

Since this version Chorda 2 plays a new voicing, measured note by note against a recorded grand (the Salamander): every key tuned within 10 cents of it, its partials, decay and unisons voiced key by key. A song saved with an earlier beta that uses Chorda 2 plays the new voicing too; a song on Chorda 1 sounds as it always did.

The sustain pedal is also a knob — Sustain, on the PEDALS card in Play and Edit — so a song can pedal without a pedal: draw it as an automation lane (right-click the knob ▸ Create automation clip; points held, up on each chord change and down just after it, is how a pianist pedals), or let a modulator move it; half pedalling works there too. The pedal the strings hear is the higher of the knob and controller 64, so a real pedal still works on top of a lane. The demo With All Your Heart is pedalled this way.

Model 2 (round 37). A Chorda you add now plays its second model: the strings reach you through the soundboard and case as two microphones over the strings hear them (the body), in a room, with the hammer's knock, the key landing on the keybed and the dampers falling back — and no two strikes of a key exactly alike. Two knobs belong to it: Room (on the RESONANCE card: how much of the room you hear; 0 % is the piano alone) and Action noise (in Edit: the knock, the keybed and the dampers; 0 % is none). A song made before keeps the first model, and sounds exactly as it did; to move its piano to model 2, load any of Chorda's presets on it (one undo step). The model is set, never automated.

Where you listen from (model 2). A modelled piano's perspective is a control, not a recording. On the MICROPHONES AND LID card: Mics moves the pair from over the hammers (0 %: close, direct, every hammer and damper heard) through the player's seat (35 %, the default) to the audience (100 %: more of the board's ring and the room, the top softened by the air); Width opens or narrows the pair (the mono sum never changes, so no setting can sound comb-filtered); Lid closes it (closed: darker, the strings covered; half; open). In Edit, the ROOM card adds Size, from a small studio to a hall (the room's decay is shown in seconds). Every one of them moves smoothly while the piano plays and can be automated. The presets use them: Close and Intimate and Pop Forward sit over the hammers, Hall Grand in the audience, Felt Piano with the lid closed, Jazz Club half open.

Changing A4, the stretch or the stiffness designs the strings again, beside the window, once the knob has settled (a second or two; the editor says designing the strings with how far it has come): the piano keeps its previous tuning until the new one is ready. An export always waits for it. These three are set, not automated.

Sectio: slicing a loop in Exemplar

Right-click a loop (WAV, AIFF, FLAC, OGG) in the sample browser and choose Slice into a new Exemplar track (or Slice into the open Exemplar), or choose Import ▸ Slice a loop… in an Exemplar's Edit view: the loop is cut into slices, one per key from C1 (MIDI 36) up, each at its own pitch — one undo step. Where it is cut: the file's own markers (a WAV's cue points, else its sampler loops; an ACID file's beats give the loop's tempo), else its transients, or a beat grid. Up to 64 slices (the status line says when more were found; a grid is made coarser to fit, and a loop with no tempo of its own uses the song's; loops longer than 5 min 49 s at 48 kHz cannot be sliced). A loop dropped on the Slices card (from the system or the browser) slices it again into that Exemplar.

The Slices card draws the loop with its slices numbered (where the numbers fit) and says how many, the tempo and how they were found:

Slicing styles are presets for Exemplar in the library (tag Slicing style): Drum Break (transients, each slice cutting the last), One-Shots (slices ring out), Stutter Grid (1/32, choked), File Markers, Vocal Chop (beats, each slice only while held; a lead), Melodic Phrase (1/8, one voice gliding). On a sliced Exemplar a style re-cuts your loop its way and plays it its way (one undo step); on an empty one it plays a few built-in sounds until you slice a loop into it. No audio ships with them.

Every slice is a zone of Exemplar's: its pitch, gain, pan, reverse, choke group and its own low-pass (Low-pass in the zone table) are the zone's, its envelope its group's. The slice points are saved with the project. Agents have slice_sample (on files inside the project's folder and the folders you allowed) and dump_slices.

The modulation effects

Eight effects move the sound in time: Concentus (chorus), Flexus (flanger), Phasis (phaser), Tremor (tremolo and auto-pan), Vibrans (vibrato), Anulus (ring modulator), Translatio (frequency shifter) and Fluctus (volume shaper). Each one is inserted on a starting preset so you hear it at once — Concentus as Juno I, Flexus as Jet plane, Phasis as Phase 90, script, Tremor as Surf amp, Vibrans as Singer, Anulus as Bell, Translatio as Slow phase, Fluctus as Pump on the quarters — named in the frame's preset well. Each control's default (what a double-click returns it to) is still the setting that changes nothing. To insert effects doing nothing instead, choose Settings ▸ Editing ▸ Effects ▸ On insert ▸ Neutral settings. Other units start where they are useful too: Fervor on Tape on the bus, Multiplex on Mastering glue, Impetus on More punch, Discrimen on More attack; Comprimo, Porta, Sibilus, Spatium and Tonus already work at their defaults; an EQ, Latitudo and the analyser start flat, and Terminus, Culmen and Tonsor — the safeties at the end of a chain — start doing nothing until something goes over. Our CLAP and VST3 versions start on the same preset in another DAW.

EffectWhat it doesWorth knowing
ConcentusThickens and widensJuno I, II, I+II are the Juno-60's chorus as its circuit sweeps; Dimension 1–4 widen without the wobble; Free voices are 1–4 voices on your own LFO. The mono check tells you whether the width survives a mono speaker.
FlexusThe jet-plane combThrough zero, the dry sound waits 5 ms so the moving copy passes through it, as two tape machines did; those 5 ms are the footer's latency, and Cantus compensates them, even when the switch is automated. Off, Flexus adds no delay. The sweep follows the LFO, how loud you play (Envelope), or a knob (Manual). The comb now draws the notches you hear.
PhasisSweeping notches2 to 12 stages (a notch for every two), feedback either way; the Phase 90 and Small Stone are presets.
TremorLevel or pan in timeTremolo, Auto-pan (a balance: the far side fades, the near side never gets louder) and Harmonic (lows against highs, like the brown amps). Square and saw shapes have a soft edge, so they never click.
VibransPitch that waversThe depth is in cents, whatever the rate; Rise lets the vibrato arrive after each note, as a singer's does.
AnulusBells, metal, robotsThe carrier is shown as a note with its cents; What a tone becomes draws the partials it makes of a test tone, its frequency (1 kHz) written on the display.
TranslatioMoves every partial by the same hertzNot a pitch shifter: harmonics stop being harmonic. Up, Down, or Both (one per side); feedback makes the barber-pole spirals. Its other sideband is more than 60 dB down from 20 Hz to 20 kHz.
FluctusSidechain pumping without a keyDraw the level over one cycle (a note value from 4/1 to 1/16): drag points (Shift: off the grid), double-click to add or remove one, drag a segment to bend it; segment shapes are the automation editor's.

In time with the song. Every LFO can follow the song (Sync, a note value with dotted and triplet values, swing where it makes sense) or run free in Hz. A synced LFO is tied to the song's position, not to when you pressed Play: start from bar 5 and it is where it would have been; a loop that wraps puts it back; offline export sounds exactly like playback. The display of each effect draws what the LFO will do over the next two cycles, with a playhead where it is now.

In another DAW. The exported CLAP and VST3 versions read the host's tempo and position, so a synced Tremor or Fluctus follows that DAW's song too.

Each one's face. The eight wear faces of their own, built around the display that shows how each moves the sound; the other channel or voice is drawn in each face's second colour. The controls work as everywhere (the value above each knob and its name under it, right-click for Modulate…, MIDI learn and automation, the keyboard and the screen reader); every colour keeps its contrast in light and dark themes, and Units wear their own faces in Settings ▸ Appearance turns them off for the plain look.

EffectIts face
ConcentusBrushed silver with rack ears, machined knobs and lamp keys for the models; each voice's delay under glass as a bucket brigade's beads, the left voices ice blue, the right ones amber
FlexusViolet lacquer, skirted knobs and bat toggles; the comb glows on a phosphor screen with a mark under every notch, and through zero the wet's line flares lime where it crosses the dry
PhasisMatte orange like a stomp box, soft knobs; the response on a fluorescent display, every notch marked in teal
TremorA seafoam vinyl amp covering with graduated dials; the shape drawn by a chart recorder's pen on paper, the other side in amber
VibransPeriwinkle behind smoked glass, value rings; the pitch in cents on a tuner's liquid-crystal panel, with its scale and a pointer at the pitch now
AnulusA walnut case — anulus is Latin for ring — with knobs that are rings; the sidebands on a lab oscilloscope in gold, each pair joined by a ring round the input
TranslatioBlack anodised rack with pointer knobs; the spectrum on a green phosphor screen, an arrow carrying each harmonic to where it lands, the right sideband in coral
FluctusCarbon weave; the shape in magenta under dark glass, the playhead in cyan with the level it rides now

Your plug-ins. When you press Scan for plug-ins (Settings ▸ Plug-ins), Cantus reads the standard folders — %LOCALAPPDATA%\Programs\Common\CLAP and …\VST3, C:\Program Files\Common Files\CLAP and …\VST3 — and any folder you added there. It also looks for new or changed files every time it starts, in the background (Look for new plug-ins at every start, on by default), and the status line names what it found — 3 new plug-ins: Odin 2, Surge XT, ValhallaSupermassive. Each file is scanned in a separate process, so a plug-in that crashes while being scanned is blocked, not fatal. Cantus's own effects are installed in your CLAP and VST3 folders too, for other DAWs.

Cribrum, the auto filter

Cribrum (Latin for a sieve) is a filter that moves in time with the song: two analog-modelled filters — the same five circuits as Cantus's synths (state-variable, transistor ladder, OTA cascade, diode ladder, Sallen-Key) — in series or in parallel, moved by an LFO, an envelope follower and a curve you draw. Insert it from the picker (it is under EQ and under Modulation); it starts on Bar sweep, a resonant ladder swept once a bar, so it is heard at once.

What passes. The window's picture is the music sieved: its spectrum as grains, lit up to what comes out of the filters and capped where it went in, under the response of both filters as they move. Drag a filter's handle across for its cutoff and up and down for its resonance (Shift is finer; a double-click puts both back). The chosen filter's handle sits on a rail that shows how far its movement can take it.

Movement. Choose the LFO, the follower or the curve on its key: its picture moves on the lane in time with the song, its own controls are beside it, and under them its depths onto the cutoff (in octaves), the resonance and the vowel. The three sources add up. The LFO runs on the song from 1/64 to eight bars (dotted and triplet) or free in Hz, in seven waves (sample & hold and a smooth drift among them), with a stereo phase. The follower hears the track or a sidechain — a kick opening a pad's filter; its range sets how much of the level moves the filter (narrow it and only each note's attack opens the filter: the envelope wah of the Cry wah preset). The curve is drawn as Fluctus's and the automation clips are (drag a point, drag a segment to bend it, double-click to add or remove one) and smoothed so a step never clicks.

The filters. Each has its model, its response (low-, band-, high-pass, notch, peak where the model gives it, and a vowel filter that talks A-E-I-O-U), 12 or 24 dB/oct, cutoff, resonance (every model sings in tune from 95 %), drive and vowel. The second filter can follow the movement, ignore it or move the other way. Spread opens the two channels' cutoffs apart.

Exact. At 0 % mix Cribrum passes the sound untouched, to the bit. A render from any bar moves the filters as the whole render does there; offline equals real time at any buffer size. Oversampling (2× or 4×) keeps a driven filter clean; its latency (a sample or two) is compensated.

Vetus, lo-fi

Vetus (Latin for old) ages a sound the way an old medium aged it. Insert it from the picker (under Distortion); it starts on Worn cassette — wow and flutter, tape hiss, the tape's band and a little wear — so it is heard at once.

Six modules, in the order a recording travels — digitised, put on a medium, worn, played back — each switched on by the lamp in its card's title (its controls answer while it is off, dimmed, so you can set a module up before you hear it):

Character is the one knob that ages everything at once: every module that is on moves towards its oldest (fewer bits, a lower rate, more wow, more crackle, a narrower band) and the label prints the year it takes the sound back to. Which modules are on decides the flavour; Character, how far back.

The window. At the centre, the object that ages — a cassette (the reels turning, the level window holding the last second of the music, the tape creased with age, a spark at the head when a tick falls) or an LP on its turntable (the groove holding the music of its last turns, off its spindle with the wow, dusty and scratched): choose it in the window's ⋯ menu, beside the colourway. Turn the object back (drag round it anticlockwise) to raise Character; double-click to make it new. Beside it, what comes out: the output's spectrum over the input's, with the converter's Nyquist (drag it across for the sample rate), the bits' noise floor (drag it up for fewer bits) and the medium's band (drag its edges).

Exact. With every module off Vetus passes the sound untouched, to the bit. The noise, the dither, the dropouts and the wear are fixed to the song: a render from any bar, offline or playing, at any buffer size, has the same crackle in the same places.

Libra, the utility

Libra (Latin for the balance) holds a channel's everyday corrections. It inserts as a wire — nothing changes until you ask.

The channels. Stereo, Swap, Mono (both summed: check a mix in mono), Left or Right on both sides, Mid or Side alone (listen to the width by itself). Invert polarity of either channel; DC removal takes an offset away; Mute.

The delays make each channel late by up to 100 ms, in whole samples — two microphones on one source brought into line to the sample. Show them in ms or in samples.

The image and the gain. The balance turns one side down, never the other up (the master's balance); width from mono (0 %) to twice as wide (200 %); the gain ±36 dB.

The window. At the centre a brass balance: its beam tilts with the balance and reads it on a fine scale, its arms reach as wide as the image, each pan holds its channel's level, delay and polarity, and the correlation runs along its foot (towards −1, the channels cancel in mono). Drag it across for the balance and up and down for the width; double-click to centre it. Beside it the tuner — the note, its cents and its frequency, its reference A (415 to 466 Hz) dragged up and down — and the goniometer.

Saltus, pitch

Saltus (Latin for the leap) moves the pitch of what it hears. Insert it from Add insert ▾ ▸ Pitch ▸ Saltus; it starts as Octave up, twelve-string (the poly engine) — or on the starting preset you chose (below).

Four modes, the legends along the top of its display:

Mono or poly. The switch at the top right of the display shows each engine's latency: Mono follows one note at a time with a voice's formants kept (2.7 ms — for playing and singing through it), Poly moves every note of a chord or a whole mix (85 ms — delay compensation covers it in a mix). The factory presets come in both.

Formants stay where they are (a voice keeps its own) or move with the pitch (the tape sped up or slowed); the throat moves them on their own.

The window. The leap: a keyboard, the note heard lit on it and named large, and an arc from it to each voice, as thick as the voice is loud, with the interval's name and the note it lands on — drag a landing to another key to change its interval. Above it a grand staff writes the last seconds as they are played, and the notes now at its right edge, each falling to its key. In Whammy the pedal is drawn beside the keyboard.

Choose its starting preset. Every unit's preset list offers Use as starting preset: its settings now become what it starts on when you insert it, in every project. Reset to factory starting preset brings back the factory's. Settings ▸ Editing ▸ Effects ▸ On insert ▸ Neutral settings still inserts every unit at its defaults.

On Tabula Saltus is a pedal with the knobs Interval, Level, Dry and Pedal.

Arca, amp and cabinet

Arca (Latin for the chest) is a valve amp and its cabinet, for guitar, bass, keys or a voice. Insert it from Add insert ▾ ▸ Distortion ▸ Arca; it starts as Clean combo.

Five amps, the legends along the top of its display: Clean, Crunch, Chime, High gain and Bass — each its circuit solved valve by valve, from the input to the speaker. None leaves only the cabinet (an impulse-response loader). The oversampling (2×, 4×, 8×) shows the latency each gives; delay compensation covers it.

The chassis shows the amp working: the valves stand in the order the signal meets them, their heaters lit, each plate reddening as hard as it works, a bar under each; the needle at the right shows the supply sagging under a loud chord. The amp's power valves, its bias and its rectifier are written beside them.

The faceplate: Gain; the tone stack's Bass, Middle and Treble; the power amp's Presence (on the chime amp, Cut, turned the other way), Master and Sag (the supply from stiff to soft); Mix and Output.

The cabinet: choose one of five, or Loaded file for your own impulse response (a WAV, AIFF, FLAC or OGG). Choose the microphone (dynamic, ribbon, condenser), drag it across the cone — bright at the cap, darker toward the edge — and back from the grille in the side view (1 to 30 cm: closer, more bass). The response the choice makes is drawn under them.

Sixteen factory presets cover every amp, the cabinet alone, keys and a voice. At its defaults (no amp, no cabinet) Arca touches no sample. As a pedal on Tabula it shows Gain, Bass, Treble and Master.

Rota, the rotary speaker

Rota (Latin for the wheel) is the rotary cabinet of Cantus's organ, for any track. Insert it from Add insert ▾ ▸ Modulation ▸ Rota; it starts as Chorale (slow).

The switch along the top: Stop (the brake parks both rotors), Slow and Fast. Each rotor ramps to its new speed at its own pace — the horn in a fraction of a second, the drum over seconds; Ramp scales those times. With Switched by notes on, each note played on the track whose notes Rota hears throws the switch, to the sample.

The display draws the cabinet cut away, the horn and the drum turning as the engine turns them (blurred at speed); the cabinet from above, where you drag the microphones round it (their angle) and away from it (their distance); and the ramps: each rotor's speed since the switch, with a dot where it is now.

The cards set the horn's and the drum's slow and fast speeds, the microphones' angle, distance and the Balance of horn against drum, the cabinet's back (open or closed) and its valve amp's Drive, and the Mix and Output. Thirteen factory presets. At its defaults (mix 0) Rota touches no sample; it has no latency.

Loquax, the vocoder

Loquax (Latin for the talkative one) makes a voice speak through another sound. Insert it on the voice's track from Add insert ▾ ▸ Pitch ▸ Loquax; it starts as Robot — the synthesiser drones C3 while the voice speaks, so you hear the voice vocoded at once, with no notes and no key.

Who speaks and who sings. The track's own sound is the voice (the modulator). The carrier is chosen on the Carrier card:

Who speaks trades them: Sidechain speaks and this track is the carrier — drums or a loop that talk. Robot drones its own note (any note, C1–C7) while the voice is louder than its gate, with no notes at all.

The display. On the left, the object the voice speaks into — the gooseneck microphone, the talk-box's tube rising from its driver, or the desk microphone on its weighted base — its on-air lamp and glow following the voice. Choose it in the ⋯ menu, under Microphone, beside the colourways: it is the insert's own (kept in the song, one step of undo) and never heard. Each factory preset suggests its object (Talk-box the tube, Radio the desk microphone, Robot the gooseneck) and loading it sets it; a preset of yours keeps the one you saved it with. A/B leaves it, and changing it never marks the preset as modified. In the middle, the bank: every band a column of cells lit by what it says now; drag across the bands to draw their levels (−24 to +12 dB, one stroke one step of undo). The formant shift is drawn along the top. On the right, the meter bridge: the voice's needle and the carrier's, the lamps for a consonant heard, the hold and the robot.

The cards: the carrier and the roles, the raw carrier let through; the bands (8 to 64, slopes of order 4, 6 or 8, their range and width); the followers (attack, release, Hold freezes the vowel); the character (formant shift in semitones, band shift, tilt, flatten, stereo spread); the consonants (the sibilance let through above its corner, always or only on consonants, and noise into the carrier on them); the synthesiser's oscillators, unison and envelope; the robot; mix and output.

Fifteen factory presets, both carriers among them (Talk-box and Sidechain carrier use the key; Classic vocoder, Choir pad and String machine the notes). In the browser each is auditioned with the source's own line as its notes and its key. At its defaults (mix 0) Loquax touches no sample; it has no latency.

In another DAW Loquax is the plug-in Cantus Loquax (CLAP and VST3): it hears the notes the host sends to it, and its carrier through the sidechain input once its own Sidechain switch is set to Sidechain input.

Fractus, the glitch sequencer

Fractus (Latin for broken) is ten effects on ten lanes of a step sequencer that follows the song, with a cloud of grains beside them. Insert it from Add insert ▾ ▸ Modulation ▸ Fractus; it starts as Last beat stutter — the last beat of every bar repeats a thirty-second, the rest of the bar untouched — so you hear it at once on drums, a loop or a strum.

The lanes, one per module: Repeat (the event's first slice plays and is caught, then replayed — each repeat higher or quieter if you ask, and with Build the stutter that halves each part), Reverse, Tape stop (or Start, up from still), Shuffle (slices from up to eight back, some backwards), Pitch jump (the classic delay-line warble, or Clean as its Style), Crush, Filter sweep, Gate, Delay throw (the echoes ring on after the step) and Cloud (the grains take over for the event). The grid is set on the Grid card: steps of 1/4 to 1/64 (triplets too), one to four bars, swing. Every start, end and jump is crossfaded, so nothing clicks: how long (Joins) and the Seed of every roll are on the Events card of the Slicing page. The delay throw's time, feedback, tone, level and ping-pong are on the Sequencer page.

Painting. On the lanes' window:

Each stroke is one step of undo. A lane's name opens the page of cards its module is set on; the four page keys under the window (Sequencer, Slicing, Colour, Grain cloud) do the same, their dots showing which lanes are painted and which sound now.

Played by notes. On the Playing card, Played by chooses Steps, Notes or Both. With notes, the notes of the track the slot listens to (its MIDI source, as for Porta and Loquax; the keyboard's track may be muted) play the lanes while they are held: the First key (C3 by default) plays Repeat, the next key Reverse, and so on up to Cloud. Notes start On the step waits for the next line of the grid; At once acts on the note's own sample. The keys that play the lanes light on the display.

The grain cloud. Its level runs beside the chain at any time (0: off): grains read from up to two seconds back, their size, density, spray, pitch (to a scale: chromatic, major, minor, pentatonic, fifths, octaves), spread and backwards share, drawn live on the scope — each grain where it reads and at its pitch, with its trail. Frozen (the lamp over the scope, or the cloud card's switch) holds what the cloud has heard for as long as it stays on.

Eighteen factory presets: stutters and builds, tape stops and starts, reverses, gates, filter sweeps, delay throws, shuffles, lo-fi fills, grain pads, a cloud to freeze, and sets played by notes (Notes: the whole kit, Notes: stutters, Steps and keys). A preset brings its lanes. At its defaults — nothing painted, the cloud at 0 — Fractus touches no sample; it has no latency.

In another DAW Fractus is the plug-in Cantus Fractus (CLAP and VST3): its lanes are saved in the plug-in's state and painted in its own window, its steps follow the host's tempo and position, and the notes the host sends play the lanes.

Caelum, the shimmer reverb

Caelum (Latin for the sky) is a reverb for spaces no building holds: its tail climbs in pitch as it rings. Insert it from Add insert ▾ ▸ Space ▸ Caelum; it starts as Octave shimmer — a wide space, a third wet, whose tail climbs an octave at a time — so a chord rises into the sky as soon as you play it.

How it climbs. Each time round the tank, a share of the tail (the Amount on the Shimmer card) is traded for the same sound shifted by the first voice (+12 semitones, an octave, at first) and, at its own level, the second voice (+7, a fifth, silent until you raise Second level). Each voice can be anything from three octaves down to three up. The trade moves the sound's energy and makes none, so the decay you set is the decay you hear, shimmer or not. Shimmer tone cuts the lows and highs inside the loop: the climbing octaves fade out at the top instead of piling up.

The window is an observatory at night:

The cards: Space (size, decay, pre-delay — free or a note value at the song's tempo, Sync — diffusion, and the tank's 8 or 16 lines: eight cost half as much), Damping (the high and low corners and their decays against the middle's), Shimmer, Shimmer tone, Motion (the lines' modulation and Bloom, the sound swelling into the space over up to four seconds), Ducking (range, threshold, attack, release) and Output (low and high cut, width, mix, output).

Ducking steps the reverb aside while the dry plays and lets it bloom in the gaps. It listens to the track itself, or to another track routed to the slot as its key (the sidechain field, as for a compressor): a vocal's shimmer that ducks under the drums.

Nineteen factory presets: octave and fifth shimmers, cathedrals, a bloom pad, Freeze swell (infinite; throw Freeze on a chord to hold it as a pad — no preset starts frozen, since a tank frozen from its first sample would hold silence), an endless ascent (its Sustain at 60 %: it keeps rising and fades into the top once you stop) and an infinite drone that build as you play, a ducked vocal shimmer, guitar ambience, pre-delays synced to the song (Eighth-note echoes, Dotted rise), a sub-octave cloud, plates, a shimmer snare and a bus haze. At its defaults (mix 0) Caelum touches no sample; it has no latency.

On Tabula's board Caelum is a pedal too (the decay, the shimmer, the first voice and the mix on its face); a board with Caelum takes a key, and the key reaches its ducking.

In another DAW Caelum is the plug-in Cantus Caelum (CLAP and VST3): its ducking hears the sidechain input once its own Sidechain switch is set to Sidechain input, and a synced pre-delay follows the host's tempo.

Vinculum, the bus compressor

Vinculum (Latin for a bond) is the compressor that glues a mix, on a bus, a group or the master. Insert it from Add insert ▾ ▸ Dynamics ▸ Vinculum; it starts as Glue (the VCA at 4:1, 30 ms attack, Auto release, −18 dB).

Three models, chosen on the faceplate's first switch:

The faceplate answers a click: a click on an engraved position chooses it, a click on a knob steps it, a click on the bat throws Compress or Limit. Its lamps are controls too: External key opens the key's list (its own input, or any strip of the desk), Auto release lights Auto (a second click gives back the release it had), Limit throws the character. The two needles show the gain reduction and the output; between them a window shows the model's gain element at work. The screen draws the transfer curve with the signal's point on it and eight seconds of reduction.

The cards: Threshold and Make-up; the Side chain — the key (also in its own field) and the High-pass that keeps the low end from pulling the detector; Valve (the vari-mu's colour); Mix (parallel compression) and Output. Fourteen factory presets cover every model: glue for a bus or the mix, drum smash and parallel drums, a vocal bus, mastering glue and a limit on each. At its defaults (the threshold at +24 dB) Vinculum touches no sample; it has no latency. As a pedal on Tabula it shows Model, Threshold, Make-up and Mix.

Lenis, the resonance suppressor

Lenis (Latin for gentle) smooths what rings: the whistles, the harsh peaks and the boxy build-ups that stand out of a sound's own spectrum, found moment by moment and taken down by a share, while the notes and the tone are left alone. Insert it from Add insert ▾ ▸ EQ ▸ Lenis (also under Dynamics); it starts as Gentle (half of each ring, the Soft curve, from 150 Hz up).

The screen is a bench analyser's: the input's spectrum as a glowing line with the cut hanging under it as a veil, a bar per frequency along the top for how much is taken away, and the focus at the foot — two gates, From and To, beyond which nothing is touched, and six numbered bands. Drag a gate left or right; drag a band's node left and right for its frequency and up and down for its amount (above the line it works harder, below it gentler, at the bottom it leaves that region alone); the wheel over a node sets its width; a double-click switches a band on or off. A band dragged while off switches on in the same step. Each drag is one step of undo.

The panel beside it: the Depth dial (drag it up or down; double-click for none) — at 100 % a ring comes down to the spectrum's own outline, at 50 % half way; the readouts of the most cut now and the latency; and the push keys: Quality (Fast, Standard, Precise: 5, 11 or 21 ms of latency at 48 kHz — Precise resolves the low mids, Fast hears sooner), Curve (Soft barely touches a peak just over the line, Hard takes every peak by its share), Listen · What is removed (only the rings, never the music) and Channels (Linked, Independent, Mid and side). A click pushes a key in; each key's menu offers automation and MIDI learn.

The cards: Detection (Sharpness — narrow rings only, or broad build-ups too — and Selectivity, how far a peak must stand out), Time (Attack, Release), Focus (the gates' frequencies), Focus band (the band chosen on its numbers or on the screen: its lamp, frequency, amount and width) and Output (Mix, Output).

The latency is compensated by the mixer like any latent insert's — a change of quality starts the analysis again (a gap of one latency) and the mixer re-aligns at once. Fourteen factory presets: harsh vocal, sibilance tamer, guitar fizz, piano ring, mix bus gentle, master polish, drum room ring, acoustic boom, podcast voice, snare ring, wide air (mid and side), strings and brass, surgical. At its defaults (depth 0) Lenis is its input delayed by its latency, to the bit. As a pedal on Tabula it shows Depth, Sharpness, Selectivity and Focus from.

Cedo, the ducker

Cedo (Latin for I make way) makes room: the bass gives way to the kick, the music to the voice, or the whole track pumps on the beat in a shape you draw. Insert it from Add insert ▾ ▸ Dynamics ▸ Cedo; it starts as Classic pump (the tempo fires it on every quarter, 9 dB down) and sounds at once, with nothing routed.

What fires it — the Trigger switch on its card (or a click on the trigger's flag on the bed):

The screen shows the shape over one cycle with what is ducked now swept in by the playhead. Drag a point to move it (it snaps to the sixteenths; Shift moves it freely), drag between points to bend the curve, double-click to add or remove a point, and right-click a point to change its segment (curve, S or held). In Sidechain the screen shows the envelope on each hit: drag the attack's, hold's and release's handles in time. Under it, the last four seconds of reduction with the hits; in Sidechain drag the threshold's dashed line up or down. Drag the level gauge beside the bellows down for a deeper duck.

Band: duck the Whole sound, only the Lows or only the Highs, split at the Crossover (drag it along the band strip) at Minimum phase (no latency) or Linear phase (no phase turn; click the split's word on the strip to switch). The Linear-phase floor sets what the linear phase's latency is sized for: Auto (the default) an octave under the crossover — the least latency it needs, changed only when you let go of the crossover — or a fixed 40, 60, 80 or 120 Hz (55.3, 36.8, 27.6 or 18.3 ms at 48 kHz), for a latency that never changes; each choice shows its latency. The mixer compensates it.

Output: Depth (∞ at its top is silence), Stereo link, Mix, Output and Listen (the output, the Delta it takes away, or the filtered Key).

Seventeen factory presets: classic, half-time, eighths and swing pumps, triplet breathe, trance gate, reverse swell, lows-only pumps at minimum and linear phase, a MIDI pump, kick ducks the bass, kick on the bass lows, hard sidechain, voice-over, podcast music bed, gentle radio and snare room (the Sidechain and MIDI ones need their key or notes chosen). At its defaults (depth 0) Cedo leaves the sound untouched, to the bit. As a pedal on Tabula it shows the trigger, depth, cycle and threshold; a keyed board's key, its beat and its notes fire it.

Purus, the real-time repair

Purus (Latin for clean) repairs a track while it plays: the mains' hum, clicks and crackle, a room's tail on a voice. Insert it from Add insert ▾ ▸ Utility ▸ Purus; it starts as Mains hum 50 Hz — the hum module on, at 50 Hz, followed as the mains drift, eight harmonics notched — so a ground loop's hum leaves at once. (The audio editor's Denoise and Declick repair a take offline; Purus does it as the song plays, on any strip.)

Three modules, each with its own switch, mix and output, each off until you switch it on:

The latency is only that of the modules switched on, and Cantus compensates it on the strip, so the track stays in time with the rest.

The window is a restoration bench:

Sixteen factory presets: mains hum at 50 or 60 Hz or found by itself, a guitar amp's hum and buzz, a live guitar DI and a bass DI (gentle), vinyl clicks, a worn record's clicks and crackle, an old cassette, a podcast in a small room, dialogue in a reverberant room, a far-microphone interview, a lecture hall, a field recording, a video call and a strong cleanup. At its defaults every module is off and Purus touches no sample.

On Tabula's board Purus is a pedal too — a de-hum pedal before the drive (the fundamental, the harmonics, the depth and the mix on its face).

In another DAW Purus is the plug-in Cantus Purus (CLAP and VST3); when a module is switched, it tells the host its new latency.

Comparatio, the reference comparator

Comparatio (Latin for a comparing, a setting side by side) puts your mix beside the records you want it to stand with. Insert it on the master's last slot from Add insert ▾ ▸ Mastering ▸ Comparatio (or Utility); it starts at its defaults — your mix heard, untouched — and measures.

Load a reference: press Load reference… on the console's foot, or drop audio files (WAV, AIFF, FLAC, Ogg) on the console. Up to eight references stay in the list, each a numbered key; press a number to compare that one, Remove to take it off. The files are read from your disk where they are — never copied into the project — and each keeps its own analysis, so its length, loudness and tone show as soon as it is read. If a file has gone missing, or changed since you loaded it, the strip says so in red: load it again.

Compare: turn the big selector (or press a lamp, or A and B in the blind drawer) to hear the reference instead of your mix, at your mix's loudness — the switch is a short crossfade you cannot hear as a click. The cards set how the reference is matched: Match Integrated (over what has played, or with Measured over The whole file, steady from the first second), Short term (following your mix section by section) or By hand, an Offset in every mode, and the Crossfade. Reference playback plays it on Its own playhead while the song plays, Locked to the song (with Ahead by), or in a Loop — drag across the overview to loop the stretch you want. Each reference keeps its own playback: compare another and its own loop and offset come back; the match, the crossfade and the display stay as they are. Only the reference compared is kept in memory: another one is read again from disk when you choose it (your mix plays meanwhile, the strip says it is being read).

The screen shows both tonal balances (the reference's from its whole file, your mix's as it plays) and your mix against the reference — above the line what your mix has more of, below what the reference has more of, the soft band how sure the comparison is. The orange line is the match EQ: what would move your mix toward the reference, as far as the Match EQ card's Amount says and never more than Most boost or cut. Press Send to Aequalis and it becomes an Aequalis just before Comparatio, in editable bands, its output gain keeping your mix as loud as it was — one step of undo; send again and that Aequalis is updated. Under it, Dynamics (the two loudness histograms, crest factor, peak to loudness, loudness range) and Stereo (correlation and width, whole and in three bands). The two decks read short-term, integrated and momentary loudness, true peak and loudness range.

Blind test: press X to hear one of the two, hidden; answer X is your mix or X is the reference; each answer lights a lamp, and the drawer gives your score and the chance of getting it by guessing. New test starts again in a new order.

Every press and drag is one step of undo; right-click the selector or a lamp for automation and MIDI learn (a footswitch for A/B). Fourteen factory presets are ways of comparing: mastering by the whole file, mixing section by section, by hand, the reference one decibel down, locked to the song, two loops, quick and smooth switching, low end in detail, a gentle and a full match EQ, an album's songs side by side, a worship mix against a live record.

Renders never hear a reference: an export, a stem, a freeze or a bounce always hears your mix, whatever the selector says. Comparatio is not a pedal on Tabula's board: a reference is the master's monitoring.

In another DAW Comparatio is the plug-in Cantus Comparatio (CLAP and VST3): Load reference… in its window, or drop a file on it; the session keeps each reference's path, size and hash (never its audio) and says when a file is missing or changed; the host's offline bounce hears your mix. Send to Aequalis is Cantus's own.

Aula, the convolution reverb

Aula (Latin for the hall) is the reverb of a real space: it plays the sound through an impulse response — a recording of how a room answers a single click — the way the room itself would. Insert it from Add insert ▾ ▸ Space ▸ Aula; it starts in Cantus's Concert hall, a quarter wet, 12 ms late (the preset Hall at 25 %).

Where the response comes from (the Source card):

The picture is the response in decibels over time, after the pre-delay, as the controls below shape it — it moves as soon as a knob does. While a new response is being prepared the display says Building the response…; the old one keeps playing until the new one is swapped in whole, without a click, and the tail of what was playing is kept.

CardControls
Length and timeStretch (50–200 %: a bigger or smaller room, lower or higher), Length (the part kept, faded at its end), Reverse (the response backwards: a swell into the sound)
EnvelopeAttack (the response faded in), Decay (100 % as recorded; lower, the tail dies sooner), Gate (where the response falls this far under its peak it ends: takes a recording's noise floor away; Off at −120 dB)
ToneLow cut and High cut (12 dB per octave, applied to the response), Damping (the highs die away faster as the tail goes on)
OutputPre-delay (0–500 ms), Width (0 mono … 200 %), Mix (dry and wet crossfaded), Level

Exact and light. Aula adds no latency: the first 64 samples of the response are played directly and the rest in partitions that grow eightfold, each one's work spread evenly over the blocks before it is needed, so a long hall costs the same in every block. It is within −130 dB of a direct convolution, gives the same bits at any buffer size, and exports, freezes and renders stems exactly as it plays. At 0 % mix it passes the sound untouched, to the bit.

In another DAW. The exported CLAP and VST3 Cantus Aula plays Cantus's own fifteen responses (a plug-in's state holds numbers, not files). Agents set its numbers with set_effect and load a file with set_impulse_response (only once you allowed files, and only inside the folders you allowed).

Fragor, the pedals

Fragor (Latin for a crash, a roar) is a pedalboard of distortion and fuzz: up to three pedals in any order, each modelled from its circuit — the knobs move the circuit's own resistors, so every position sounds as the pedal does. Insert it from Add insert ▾ ▸ Distortion ▸ Fragor; it starts with the green overdrive on the board, pushing gently.

The pedals:

The board: drag a pedal to another slot (the two swap places); click an empty slot to add a pedal; right-click a pedal to put another in its place or take it off the board. The footswitch under each pedal bypasses it without a click. Right-click a knob for its menu (automate, modulate, learn a controller, reset); double-click resets it.

Over the board, the display shows one cycle of the note going in (faint) and coming out at the level playing, and its harmonics — odd ones solid, even ones hatched, the dominant ones marked — with the total distortion. The valve amp's tubes glow a little brighter while sound plays.

Beside it: Gate (before the pedals: off at −90 dB), Input (the level into the pedals), Mix, Output, Quality (oversampling 2×, 4× or 8×; more removes more aliasing and costs more) and Stereo / Mono (one circuit for both channels, as a real pedal).

Its look: the window's ⋯ menu offers three colourways — Valve (the amber of a valve amp at night, its face), Cream (cream card and brass) and Steel (steel with one signal red). The choice is yours for every Fragor and is kept between sessions; it never changes the sound.

History: the five circuits are those of the Ibanez TS808 Tube Screamer, the ProCo RAT, the Boss DS-1, the Dallas Arbiter Fuzz Face and the Electro-Harmonix Big Muff π, modelled from their published schematics; the names and the pictures in Cantus are its own.

Tabula, the pedalboard

Tabula (Latin for the board) is a pedalboard of Cantus's own effects: up to sixteen of them as pedals, in any order, with an optional A/B split and eight macros. Every pedal is the effect itself, so a board of one pedal sounds exactly as that effect on its own. Insert it from Add insert ▾ ▸ Guitar and amp ▸ Tabula; it starts on a guitar rig — the green overdrive low, a chorus, a delay and a spring reverb — so it is heard at once.

The board. The pedals stand on a dark board, cables carrying the signal between them and glowing as loud as it is at each point. Turn a pedal's knobs where they are; click its footswitch to switch it on or off (a 10 ms crossfade that never changes the latency). Drag a pedal to move it (drop it between two others, or onto lane A or B to split the chain). Double-click it, or choose Open its full editor, to open the effect's own window in place of the board; Back to the board returns. With many pedals the board wraps into rows, the cable running between them.

Adding and taking off. Add pedal at the end of the chain opens the picker: the shelves down its left, the shelf's effects as small pedals; click one and it joins the board, sounding at once. Under the board, Its place on the board sets the chosen pedal's Size (mini, standard, wide, double, round), its Place in the chain (Before, A, B, After) and its footswitch, and Remove takes it off.

The split. A pedal on lane A or B opens the split: the splitter box feeds both lanes, the mixer box joins them, its knob the A/B balance (left: A only, right: B only). The shorter lane waits for the longer one, so the two never smear (no comb filtering).

Macros. Eight knobs along the top. Press Learn, touch a macro, then touch a pedal's knob: the knob is mapped, from where it stands to the far end of its range, and wears a ring. A macro moves everything mapped to it; it wins over automation of the same knob. Randomise gives every pedal's knobs new values within musical ranges (one undo step).

Under the board the chosen pedal's main controls (each with its menu: automate, modulate, learn a controller), and the board's Input, Mix and Output.

The keyboard. Click the board, then: Tab and Shift+Tab choose a pedal; Space its footswitch; Enter its full editor; Delete takes it off; Ctrl+← and Ctrl+→ move it; ← and → choose one of its knobs and ↑ and ↓ turn it (with Shift, finely).

Presets. Each of Tabula's presets is a whole board: Guitar rig, Blues crunch, Ambient swells, Fuzz lead, Phaser funk, Wet, dry, wet, two for bass, two for synths, two for drums and two for voice. Your own presets keep the whole board too.

In another DAW, the exported Tabula shows the same window; the host automates the macros, the input, mix, output and the A/B balance, and the board is saved in its state.

Lapsus, the time shaper

Lapsus (Latin for a slip) bends time and level to the beat, as FL Studio's Gross Beat does: tape stops and starts, half time, stutters and rolls, reverse, scratches, gates and pumps. It holds two banks of thirty-six patterns — time patterns and volume patterns — chosen apart and played together. Insert it from the picker (under Modulation); it starts on Half time (every half bar's first beat an octave down), so it is heard at once.

The timeline. The time curve is drawn above, the level below, over the longer of the two patterns' cycles. The time curve's height is how far behind the music it reads: flat at the top is the music as it comes; a line falling at the diagonal is the tape stopped; half that slope is half speed; steeper is reverse; a held step is a repeat. What lies before the cycle's start is never read (Before the cycle: silence). While it plays, the playhead crosses both lanes and the slip — the dashed line to the ruler — shows where the music is read from. Drag a point (Shift: off the grid), double-click to add or remove one, drag a line to bend it; choose Pencil to paint steps a grid cell at a time. Every edit is one step of undo.

The banks. The pads show the time bank or the volume bank (the switch on top names what each plays). A click plays a pattern; a right-click gives a slot the factory's pattern back. Notes play them too: from Time note (C4) up for the time bank, from Volume note (C1) up for the volume bank — switched until another note, or only while held (Each note), and from their beginning with Restart on note. Route a track's notes to it from the insert's menu.

Exact. With both banks on Through (pattern 1) Lapsus passes the sound untouched, to the bit. Jumps of the time curve are crossfaded (Crossfade, 2 ms), the level glides (Smooth); a tape stop ends on the bar's own sample at any tempo, and a render from any bar is the whole render there. Ahead lets the time curve read the future at the cost of that much latency (compensated).

Delay, the stereo delay

Delay is Cantus's stereo echo: a time up to two seconds or a note value that follows the tempo, feedback, a tone filter in the feedback path so each repeat is darker than the one before, and ping-pong between the sides. Insert it from Add insert ▾ ▸ Space ▸ Delay; a click on its insert opens its window (the mini view under the insert keeps the time, the feedback, the tone and the dry/wet at hand).

The echoes are the window's picture and its control: the left side above the centre line, the right below; the sound at 0, then each repeat at its time, its height its level against the first, its colour cooling from the face's coral to rust as the tone takes its highs. With Ping-pong on, the first repeat is on the left and they cross from side to side; off, each side repeats itself. Drag on the picture: across for the time (the first repeat follows the pointer), up and down for the feedback (the dashed level lines follow it) — one step of undo. Shift is fine; a double-click goes back to 375 ms and 40 %. While sound plays, each repeat lights as the sound reaches it; at rest nothing moves.

CardControls
What is heardthe repeats to −60 dB, the tail they make, what the tone takes from each repeat at 8 kHz, the side of the first
TimeTime (1 ms … 2 s), or the note value while synced; Tempo: Free · Sync, the note value (1/64 … 4/1, straight, dotted ·, triplet T, each listed with its time at the song's tempo now) and what the time is now
RepeatsFeedback (0–95 %), Tone (200 Hz … 20 kHz, in the feedback path)
OutputPing-pong, Mix (the repeats against the dry sound)

A change of time glides, so a sweep bends the pitch as tape does.

Synced to the tempo (round 40). Turn Sync on and the time is a note value at the song's tempo, following every tempo change and the tempo track: inside a tempo ramp the repeats slide with it, as a tape echo whose speed follows the ramp; at a sudden tempo change or a new note value they cross-fade to the new time in 50 ms — no click, no sweep. Sync picks the note nearest the time you had; choosing a note in the list turns the sync on; Free keeps the time it was playing in milliseconds (so a note value set once is Sync, then Free). While synced the big knob turns the note value, and a drag across the echoes moves to the note nearest the pointer. A song saved before round 40 plays as it did. Long values take more memory (a whole note about 5 MB, 4/1 about 18 MB at 48 kHz); a dotted 4/1 below 30 BPM is longer than the line and is held at 48 s (the window says so). The presets named by a note value (Eighth echo, Dotted eighth, Quarter throw, Trap vocal, Sixteenth bounce, Triplet) follow the tempo.

In another DAW. Delay is also a CLAP and VST3 plug-in, Cantus Delay, with this window; synced, it follows that DAW's tempo.

Plug-ins from other makers: installing and using them

Cantus plays instruments and effects made by others in the CLAP and VST3 formats, 64-bit. VST2 plug-ins (a lone .dll) and Apple's AU are not supported: when a plug-in offers several formats, install its CLAP or VST3 version.

On a Mac plug-ins live in /Library/Audio/Plug-Ins/CLAP and …/VST3 (for every user) or the same folders in your home's Library (for you); Cantus reads both. A plug-in there is a bundle (Name.clap, Name.vst3) and its editor opens in a window of its own that stays above Cantus while you work in it.

  1. Install it with its maker's installer, downloaded from the maker's own site. Leave it in the standard folder the installer proposes (C:\Program Files\Common Files\VST3 or …\CLAP).
  2. Let Cantus find it. Restart Cantus: at start it looks for new plug-ins by itself and says what it found. Without restarting, choose OPTIONS ▸ Scan for plug-ins (the top bar's menus, beside the logo), or press Scan for plug-ins on the picker's empty CLAP or VST3 shelf. The settings live in OPTIONS ▸ Plug-ins and folders… (Ctrl+, opens Settings; its Plug-ins page is in the sidebar; the picker's Find more plug-ins… goes there too). Later scans read only new and changed files. For a plug-in in another folder, first add the folder under Folders (Browse…, then Add) (Rescan everything reads every file again, blocked ones included). What it found is listed under Found; a file that crashed while being read is listed under Blocked, with Try again.
  3. An instrument: press the + above the track headers (a new instrument track), or click a track's instrument chip to swap its instrument. In the device picker, search by name or choose the CLAP or VST3 shelf, and press Enter.
  4. An effect: click an empty insert slot — in a mixer strip or in the inspector — and pick it the same way.
  5. Its own window, or Cantus's view. A plug-in opens its own editor when you insert or pick it. Cantus view shows its parameters instead, drawn as Cantus's own devices are: grouped by the plug-in's pages, a search field (by name or by page), the readings it reports (a gain reduction, a level) as meters. Switch between them with Plug-in window | Cantus view at the top of Cantus's view, under Editor in the insert's menu (right-click a slot, or its ⋯), or the Editor row of an instrument's inspector. As default (Use as default for this plug-in) makes the view showing the one this plug-in opens in from then on, in every project; click it again to go back to its own editor. The choice is yours, kept in your settings — not in the project — by the plug-in itself, so an update keeps it. A plug-in with no editor of its own has only Cantus's view. Switching never touches its sound, its state or its automation. Its values can be automated, modulated and mapped to a MIDI controller like those of Cantus's own devices, from either view.

Each plug-in runs in a helper process by default, so one that crashes stops alone and Cantus keeps playing (see When a plug-in crashes below).

Free plug-ins worth trying (each from its maker's site; Cantus has no tie with them):

Plug-inWhat it isWhere
Surge XT (CLAP, VST3)A large open-source synthesizer: pads, basses, leads, thousands of presetshttps://surge-synthesizer.github.io/
VitalA wavetable synthesizer; its free edition has plenty of presetshttps://vital.audio/
Odin 2 (CLAP, VST3)A thorough virtual-analogue synthesizerhttps://thewavewarden.com/pages/odin-2
DexedA Yamaha DX7 clone: FM electric pianos and basseshttps://asb2m10.github.io/dexed/
Decent SamplerA sample-library player; hundreds of free instruments for it, Pianobook's among themhttps://www.decentsamples.com/product/decent-sampler-plugin/, https://www.pianobook.co.uk/
Valhalla SupermassiveHuge, spacious reverbs and delayshttps://valhalladsp.com/shop/reverb/valhalla-supermassive/
TDR Nova, TDR KotelnikovA dynamic EQ and a mastering compressorhttps://www.tokyodawn.net/tdr-nova/
ChowTapeModelAnalogue tape saturationhttps://chowdsp.com/products.html
Airwindows Consolidated (CLAP, VST3)Hundreds of Airwindows effects in one plug-inhttps://www.airwindows.com/consolidated/
Dragonfly ReverbHalls, plates and rooms; open sourcehttps://michaelwillis.github.io/dragonfly-reverb/
Kilohearts EssentialsA toolbox of small effects: chorus, distortion, filters and morehttps://kilohearts.com/products/kilohearts_essentials
Youlean Loudness MeterA loudness meter in LUFShttps://youlean.co/youlean-loudness-meter/

Tonus: pitch correction for a voice

Tonus (in the picker's Pitch · vocal shelf) tunes a sung voice as it comes in — for the hard, stepped effect, for a natural correction, or for monitoring while you record (it adds 2.7 ms, reported to delay compensation).

  1. Put Tonus on the voice's track. Choose the key and the scale; on the keyboard, click a note to remove it (×) or bring it back (●), right-click to leave it alone (○).
  2. Speed: 0 ms (Hard) jumps from note to note; 50–100 ms corrects and keeps the voice's life. Amount sets how much is applied.
  3. To tune to what you play instead: MIDI notes (the notes held on a track, in any octave) or MIDI melody (the voice sings exactly the newest note) — choose the track under the switch; its clips and what you play on it live both count.
  4. More holds humanise (long notes keep their movement), flex (small deviations left alone), vibrato (the singer's, removed or doubled; or a created one), the formants and the throat (brighter or darker without changing the pitch), transposition and the reference A4.

The display shows what you sang (dim) and what comes out (bright), and the note you are singing with its distance in cents. Presets: Natural, Tight, Hard tune, Throat up, Throat down, MIDI lead and more. Details: docs/plugins/tonus.md.

Dynamics and analysis: Tonsor, Impetus, Discrimen, Speculum

Four effects shape a sound's peaks and attacks and show what is in it. Impetus starts on More punch and Discrimen on More attack; Tonsor, a safety, starts at 0 dBFS and changes nothing until a peak goes over, and Speculum only looks.

EffectWhat it doesWorth knowing
TonsorA clipper: cuts the peaks offCeiling, Drive, a Knee and three shoulders (Hard, Soft, Analog). Oversampling up to 16× keeps the harmonics it makes from folding back; True peak holds the ceiling between the samples too (about 1 ms of look-ahead), so a streaming service's −1 dBTP is really −1. Gain match takes the drive back out, so louder never wins. The meter shows how much (the deepest reduction), how often (the share of samples clipped in the last second), the overs between the samples in the input, and the output's true peak.
ImpetusMore or less attack and sustainWorks on level differences, not a threshold: a quiet hit and a loud one are shaped alike. Speed is how long an attack lasts, Length how long a tail counts as one. Highs or Lows shape only one side of a crossover (snap without the low end pumping, or a kick's thump); a sidechain lets another track's hits drive it. A hit through it draws what it does before you play.
DiscrimenSplits each hit into its transient and its bodyEach part has its own gain, a one-knob compressor and a tilt; Listen plays one part alone. Left alone the two add up to the input exactly. Zero latency costs nothing; Look-ahead (2 ms, compensated) catches every attack from its first sample.
SpeculumAn analyser that never touches the soundTabs: Overview, Spectrum (averaging, slope, peak hold, mid/side/left/right), Spectrogram, Scope (free, triggered, or on the beat), Vector (and the correlation: under 0 means trouble in mono), Loudness (momentary, short-term and integrated LUFS, range, true peak, against a target — the EBU R128 readings). ⋯ ▸ Dock in the workspace makes it a tab you can resize, split or tear out; dock several and watch them at once.

Pages: docs/plugins/tonsor.md, impetus.md, discrimen.md, speculum.md.

Each dynamics unit's face

The dynamics effects wear faces of their own, each built around the display that tells its story — how much it takes away. The controls work as everywhere (the label and value under each knob, right-click for Modulate…, MIDI learn and automation, the keyboard and the screen reader); every colour keeps its contrast in light and dark themes, and Units wear their own faces in Settings ▸ Appearance turns them off for the plain look.

EffectIts face
ComprimoA black anodised rack face with rack ears, amber legends and skirted knobs; a needle meter reads the gain reduction, and the history and transfer are drawn on cream chart paper
TerminusFlat graphite with a red line under the name — the ceiling; the output in ice blue and what it takes in red behind dark glass, value rings for knobs
MultiplexSmoked violet glass; the bands glow in the face's violet, cyan and the hues between on a phosphor screen
SibilusBrushed silver with screws and vents; the sibilance glows ice-cyan, machined knobs
PortaA mahogany door — porta is Latin for door — with green for open and amber for shut on a liquid-crystal panel, soft knobs
CulmenDeep sapphire lacquer with rack ears, knobs ringed with LEDs; the bands and loudness on a fluorescent display
TonsorCarbon weave; the transfer curve drawn by an orange beam on an oscilloscope's graticule
ImpetusBlack vinyl amp covering like a drum amplifier, pointer knobs; attack in lime, sustain in violet
DiscrimenA printed lab card with graduated dials; the transient in magenta and the body in teal, as a two-beam scope's channels

Each tone, space and analysis unit's face

The tone, space and analysis effects wear faces of their own too, each built around its display. The controls work as everywhere (the label and value under each knob, right-click for Modulate…, MIDI learn and automation, the keyboard and the screen reader). Every colour keeps its contrast in light and dark themes, and Units wear their own faces in Settings ▸ Appearance turns them off for the plain look.

EffectIts face
AequalisBlue anodised aluminium, like a precision instrument: the curve in cyan behind smoked glass, the analyser in amber behind it, small capped knobs, an engraved name, values in even-width figures
FervorWarm brown vinyl amp covering with screws and vents, chicken-head knobs and bat toggles, the name on a badge: the curve glows in an amber fluorescent window, and the harmonics show even in amber and odd in filament yellow
SpatiumDeep indigo lacquer, a hall at night: the decay and the tail glow on a blue phosphor screen, the early reflections in cold cyan, with soft knobs and lamp keys
LatitudoBrushed steel in a rack with screws and ears, like a broadcast phase meter: the stereo field and the tone are drawn on an oscilloscope's graticule, the mid in green and the side in orange, with graduated dials and rockers
SpeculumGrey-green enamel, like a bench analyser: the spectrum, the scope and the loudness are drawn on a backlit liquid-crystal panel, the held peak in red, with value rings and narrow capitals
TonusMagenta behind smoked glass, like a vocal booth: the pitch is drawn on chart paper, what comes out in magenta and what was sung in teal, with soft knobs and lamp keys

When a plug-in crashes

A plug-in made by someone else — a CLAP or VST3 effect or instrument — can crash or hang. Cantus runs those plug-ins in a helper process of their own (cantus-sandbox), so when one fails, the helper stops and Cantus keeps playing:

Where a plug-in runs. Settings ▸ Plug-ins ▸ Where third-party plug-ins run sets the default for this computer: Together (one helper for all of them; the default), By maker (one per maker), Individually (a helper each: a crash stops only that plug-in) or Within Cantus (no helper: the fastest, and a crash closes Cantus). Each plug-in can differ: right-click its slot (or open Runs on a hosted instrument's row in the inspector) and choose. The choice is saved with the project; changing it moves the plug-in at once, with its sound. A plug-in that crashes three times in a session is offered Run individually on its card — by name, when the card lists the plug-ins it took down with it, which are reloaded where they were. Only the plug-in that crashed is counted. The plug-in manager lists the plug-ins that ever stopped on this computer (Forget clears one). Cantus's own effects and instruments always run within.

What it costs. A plug-in in a helper adds a round trip per audio block: a few microseconds while plug-ins follow one another, a few tens of microseconds after a pause (ADR 0063 has the measurements). Point at a hosted slot to see where it runs and its CPU; the status bar counts Plug-ins late each time a helper did not answer in time (the sound went on without the plug-in for a moment; a larger buffer gives it more time). Exports, stems, freezes and bounces wait for every block, so they are always exactly what the plug-in makes; if a plug-in stops working during one, Cantus says so instead of calling it done, and Stop ends it even while a plug-in hangs.

11. Projects, autosave and settings

Templates and the start page

File ▸ New from template… (or Start page in the command palette) opens the start page: at its left a card for each template, at its right your recent projects (the last twelve you opened or saved; one that is no longer where it was says so). Click a card for a new, untitled project made from it — Cantus asks first if the open one has unsaved changes, as New does; click a recent project to open it.

Templates live in Cantus's data folder (%LOCALAPPDATA%\Cantus\templates), project templates in projects, track templates in tracks; each is a .cantus file, so a template from an older version opens as any project does.

Settings (Ctrl+,, or the Options menu, which also opens each page directly; View ▸ Settings… still works). The window has a sidebar of pages, as FL's options, and opens on the page you left, in the next session too: General (the verse at start, on or off, and the Spanish Bible), Audio (driver, devices, rate, buffer, latency, the record offset, and the Audio cores at work — the audio thread and the cores helping it, 1 for the audio thread alone — each count shown with an engine and a model it is known from, drawn at the right of the list and large under the field: 4 · V4 · Fulvia, 12 · V12 · Enzo, more than 16 a turbo-ramjet; the line under it says how many are at work now, counted the same way; the real-time priority each audio thread was given is the last line of the band), MIDI, Appearance (language — English or Español —, the look's family and its mode — see the look —, interface size, the cable style, Units wear their own faces — on or off, see Each unit's face —, colours, the taskbar icon's colours, keyboard shortcuts), Saving, Sample cache (its size limit — 16 GB by default; when full, the least recently used samples go first — and Clear cache; it lives in %LOCALAPPDATA%\Cantus\cache\samples, or in a portable copy's data folder), Updates, Plug-ins and AI (see AI agents below).

AI agents (MCP)

Cantus can be driven by an AI assistant you already use — Claude Code, Claude Desktop, Cursor, VS Code, Gemini CLI, LM Studio — through the Model Context Protocol (MCP). The assistant brings its model; Cantus brings the tools: tracks, notes, the mixer, effects and their parameters, automation, export. Every change an agent makes is one step in the undo history, named AI: … (AI: write the verse), and Ctrl+Z takes it back like any of yours. Cantus sends nothing to any AI service itself; what your assistant sends to its own is its business (read its privacy terms).

Turn it on. Settings ▸ AI ▸ Allow external agents: On. Cantus opens a door on this computer only — http://127.0.0.1:39571/mcp, the port changeable below it — protected by a token kept in your settings folder, readable by you alone. Copy the token puts it on the clipboard; Make a new token disconnects every agent (set each client up again).

Connect a client. Copy setup for has a button per client; each copies what that client wants, with its own token, and says where it goes:

A client that starts a program instead of an address uses the bridge too; when Cantus is not running, or does not allow external agents, it answers with a message saying so.

What an agent may do. Reading the project and the transport (play, stop, locate) run at once. Edits ask you first: a card at the bottom right says what the agent wants to do — Claude Code wants to: …, what it would touch — with Allow, Allow for this session (that one tool, from that agent, until it disconnects) and Deny; with no answer within two minutes nothing is done. Edits: Without asking lets edits run at once. Removing or replacing your material, undoing and files always ask, whatever that setting says. Files are off until you choose Ask each time, and then reach only the project's folder and the folders you add below it. Never allow refuses a tool by its name. Names and notes inside a project are treated as data, never as instructions.

Watch and undo. While an agent is connected, a chip in the top bar shows its name (its light turns blue while one of its calls waits for you). A click opens the AI activity: the agents, each with Undo all of its changes — it undoes the agent's steps from the most recent and stops at the first one that is not the agent's, saying which — and Disconnect; the jobs at work (long renders), with their progress; and every call and what became of it, refusals included. A right-click on the chip opens Settings ▸ AI.

One switch. AI area: Off refuses every agent, closes the door and hides the area; Cantus is a complete DAW without it.

12. Keyboard shortcuts

Help ▸ Keyboard shortcuts lists every shortcut and lets you change them, or load the FL Studio, Ableton Live or Bitwig Studio presets. The defaults:

KeysAction
SpacePlay / pause (also stops a recording)
RRecord
HomeGo to start
Ctrl+L · Ctrl+MLoop on/off · metronome on/off
L · Shift+TPattern / song mode · tap the tempo
Alt+R · Alt+C · Alt+P · Alt+TArm the selected track · count-in · punch · show takes
Ctrl+Shift+AAdd an audio track
Ctrl+N · Ctrl+O · Ctrl+S · Ctrl+Shift+SNew · open · save · save as
Ctrl+Shift+NDemo songs
Ctrl+Shift+EExport audio
Ctrl+Z · Ctrl+YUndo · redo
Ctrl+X · Ctrl+C · Ctrl+V · Ctrl+DCut · copy · paste · duplicate
Ctrl+A · Delete · EscSelect all · delete · clear the selection
Ctrl+E · Ctrl+J · 0 · F2Split clip · glue · mute clip · rename
Left Right (Shift: a bar)Nudge
Up Down (Shift: an octave)Transpose notes
QQuantize
P B E D T C Y or 1–7Piano roll tools; in the arrangement E P C D T (or 3 1 6 4 5) pick Select, Draw, Slice, Erase and Mute, 8 Zoom, and B or 2 is the paint tool
Shift+3The grid's triplets
Shift+L · ZLoop the time selection · zoom to it
Alt+AAutomate the last parameter you moved
Alt+Shift+ALink the last parameter you moved to a MIDI controller
Ctrl+Shift+MMIDI learn mode
Ctrl+Shift+KShow the controller mappings
F5 F6 F7 F9 · Alt+F8 · IArrangement, channel rack, piano roll, mixer · browser · inspector
Alt+F5Clip browser
F8 · F12The selected track's instrument · close floating devices
M · S · Alt+Left/Alt+RightMixer: mute · solo · move the strip
Alt+F9The routing matrix
Alt+U · Alt+L · Shift+Alt+L · Alt+NMake unique · link the selection to this clip · select linked copies · link to…
Ctrl+TTyping keyboard plays notes
Ctrl+Plus · Ctrl+Minus · Ctrl+0Interface zoom
Ctrl+,Settings
Ctrl+Shift+P · Alt+HThe command palette · the undo history
Ctrl+FFind a preset for the selected track's instrument
Ctrl+F6 · Ctrl+Shift+F6The next area · the area before
Ctrl+Alt+Enter · Ctrl+Alt+CMaximise the area you work in (again: give the others back) · collapse or expand it
Shift+F10The menu of the selection (in the mixer: the selected strip's route menu)
Alt+S · Alt+Shift+SThe score under the piano roll · the Score pane
Tab · Enter · Shift+EnterSession ⇄ arrangement · launch the slot · launch the scene
Ctrl+Enter · Ctrl+Shift+Enter · Alt+EnterStop the track's clip · stop all · record into the slot
Ctrl+I · Ctrl+Alt+I · Alt+B · Ctrl+Alt+RInsert scene · capture scene · back to arrangement · arrangement record

On a Mac Cmd (⌘) takes the place of Ctrl and Option (⌥) of Alt, and the macOS preset is the default: the table's keys, except that the metronome is Cmd+K (or K while the typing keyboard is off; ⌘M minimises the window, as in every Mac application) and redo is Cmd+Shift+Z. The menu bar at the top of the screen holds Cantus's menus with their shortcuts, which follow any change you make here. On a trackpad, pinch to zoom an editor's time and swipe with two fingers to scroll it; Cmd with a swipe zooms too.

13. Privacy and updates

Cantus has no telemetry. An installed or portable copy asks GitHub once a day whether a newer version is out — an anonymous request that carries no version, machine or user information — and asks you before downloading anything; Settings ▸ Updates ▸ Check daily turns it off.

When you say yes, Cantus downloads the new version's file for your system, checks its SHA-256 against the release (and, on Windows, the publisher's signature when the build is signed) and applies it when you close Cantus — or now, with Restart and update:

A portable copy, a copy from a Linux distribution's packages or Flatpak, and a system no release has a file for are told about the new version and shown its page. Sound packs are downloaded only when you click Install. Everything else stays on your computer. The details are in PRIVACY.md; the licences of everything Cantus is built from are in THIRD-PARTY-NOTICES.md (also in Help ▸ About Cantus).

14. Known issues and limitations

This is a beta. What we know about:

15. Report a bug

Help ▸ Report a bug… opens the bug form on GitHub (you need a free GitHub account). Say what you did, what you expected and what happened; give your Cantus version (Help ▸ About Cantus), Windows version, audio interface, driver, rate and buffer (all in the audio chip); attach a small project that shows the problem if you can. Tell us about your audio interface even when it works, with the hardware report, and send ideas with Help ▸ Suggest a feature…. English or Spanish, both are welcome.

Cantus is made by Luis Ángel Almazán — lalmazan.com — and is free software under the GNU GPL, version 3 or later.