Skip to content

Sound

Sound is the production canvas. It combines the device graph with the tools that bring recorded or managed audio into the same source-owned project.

The Browser and Devices view expose the generated instrument and effect catalog: built-in synths, FAUST voices and processors, samplers, SoundFonts, MIDI/external outputs, parameter ranges, units, automation support, macros, latency, tails, and CPU class. Factory presets are code-ready parameter sets rather than opaque binary state.

Device edits write bounded literals back into TypeScript when the selected expression is statically safe. Computed expressions remain source-only.

The Mixer projects tracks, return buses, sends, and the master. Track effects and routing live in code; fader, pan, mute, and solo remain live performance state. The Audio Session maintains one ownership, clock, route, and latency/PDC model across playback, recording, external returns, and export.

The Library stores immutable content-addressed samples, recordings, SoundFonts, stems, freezes, and bounces with metadata sidecars. Metadata includes analysis, waveform peaks, BPM/key confidence, LUFS/true peak, license, collections, project pins, ownership, and lineage.

Audio regions remain source-backed clips with stable ids. Clip Detail can edit start, length, fades, gain, channels, stretch, pitch, and warp markers. Crop and consolidate derive new immutable assets and can relink the source region without overwriting the original bytes.

The Record view supports MIDI capture, exact-frame mono/stereo audio punches, pre-roll/count-in, safe monitoring, loop-pass TakeStacks, recovery, and calibrated external-return printing.

Vocal Notebook keeps lyrics, beat timing, pronunciation, harmony, ad-libs, performance direction, and preferred managed takes in ordinary vocals.ts source. Stable lyric and line ids connect notebook rows, source selection, and agent edits.

A reference track becomes a measured feature profile for live and offline comparison. The Master Scope shows loudness, true peak, spectrum, and musical features; the MCP collaborator can call compare_to_reference over matching bar windows.

Export and loudness covers full mixes, managed bounces, six stem topologies, explicit sample rate/tails/bit depth/dither, render provenance, source-linked freeze, and hardware-print requirements.

Audio Setup owns machine-local facts that never enter song source:

  • input/output selection and verified physical output identity;
  • sample-rate/channel/browser-processing probes;
  • acoustic loopback calibration and manual trim;
  • safe headphone-monitoring approval;
  • logical external-return bindings;
  • MIDI Clock leader/follower state, jitter, drift, and Song Position;
  • graph latency and compensation evidence.

Device permission, route choice, calibration, monitoring approval, file selection, and record arming remain human gestures. The resulting source, assets, diagnostics, and renders are available to the MCP collaborator.