Measured, not claimed

Faster to start.
Quieter at rest.
Heavier on memory.

Cantus, REAPER and FL Studio, five starts each on the same quiet laptop (Cantus beta 14 on 11 November 2026, the others on 9 November), watched by the same script — each beside what it is fair to compare it with. Where Cantus loses, it says so here first.

  • Ryzen 7 7735HS · 31 GB
  • Windows 11
  • 2026-11-09 · 2026-11-11
  • Cantus beta 14 · REAPER 7.85 · FL Studio 24.1.2
  • Not measured again on beta 16 yet
Won

Window ready

seconds · lower is better

An empty song

  • Cantus0.76
  • REAPER1.76

A beat starter

  • Cantus0.77
  • FL Studio2.49

2.3× sooner than REAPER, 3.2× sooner than FL Studio.

Won

Quiet

seconds · lower is better

An empty song

  • Cantus1.5
  • REAPER2.0

A beat starter

  • Cantus1.5
  • FL Studio5.8

Done working after it opens: under a tenth of a core, sooner.

Won

At rest

% of a core · lower is better

An empty song

  • Cantus1.1–1.5
  • REAPER2.3–3.7

A beat starter

  • Cantus1.7–2.0
  • FL Studio6.6–6.7

Every channel that hears silence stops running, to the same bits.

Lost

Memory

MB · lower is better

An empty song

  • Cantus140
  • REAPER62

A beat starter

  • Cantus143
  • FL Studio133

Round 39 made the engine's reservation grow with the song: 237 MB became 140 on an empty song (−41 %), 239 became 143 on a starter. About a fifth of what remains is Cantus's own live memory; the rest — the graphics driver's buffers, thread stacks, the audio drivers' heaps — is the work ahead. Still lost, by 78 MB and 10 MB.

Features, against twelve DAWs

97 rows, each from the vendor's own manual

FL Studio, Live, Bitwig, Logic, Cubase, Pro Tools, Studio Pro, REAPER and four free DAWs. Checked 2026-11-04; re-run every release.

22 rows where a leader has more (missing or partial in Cantus, 2026-11-04)

  • Clip groups6 leaders
  • AU hosting on macOS5 leaders
  • Video track5 leaders
  • ARA 2, whole5 leaders
  • Surround / Atmos4 leaders
  • Granular / spectral sampler4 leaders

3 rows no commercial leader has

  • An agent interface over MCP, each change one undo steponly
  • Open source, at this classonly
  • Offline equals real time, the same bits on three systemstested

How it was measured

  • Five measured starts of each program after one warm-up, each closed politely, by tools/scorecard/startup.ps1.
  • Window ready: from the start to a main window that is visible and answers (not a splash, not a dialog).
  • Quiet: the program uses under a tenth of one core. At rest: one core's share from 30 s to 40 s after the start, by thread time and by cycles; both are published.
  • Like against like: Cantus on an empty song beside REAPER's empty project, on its Beat starter beside FL Studio's template.
  • AMD Ryzen 7 7735HS, 31 GB, Windows 11, High performance plan, nothing else open. FL Studio is a trademark of Image-Line; REAPER of Cockos.

Capacity, honestly

On one core Cantus sustains 120 of its compressors or 40 synth voices at 64 samples — its own offline estimate. The live DAWbench, the same session rebuilt in every host on a real interface, comes next; until it is made, no capacity comparison is claimed.

The script, the raw numbers and the whole matrix are in the repository. Run it on your machine.How it was measured →All 97 rows →