Audio Scene Canvas and Assistant Scene Rack (AssistantApp)

Inside the project

Prepared with the project’s development agent from source and retained evidence. This is a project-owner assessment, not an independent customer review. Dated checks were not rerun for this website.

Audio Scene Canvas is a node canvas inside AssistantApp for assembling an audio scene out of the app's own audio modules — NoiseColour character, the Acoustician masker, Semantic Monster and Synth, vocoders, spectral EQ, and the shipped instrument and MIDI-effect extensions — then exploring it with macros, MIDI and expressive controllers. It is deliberately not Max/MSP: where open-ended patching gives you everything and promises nothing, this canvas ships a curated set of modules whose execution lane, real-time safety, export eligibility and readiness on THIS machine are all known to the host. The practical payoff is that the canvas can tell you what will actually play — here, and in Logic — before you invest hours in a scene. Two verdicts are kept deliberately apart: 'plays here?' (is an implementation registered and the plug-in present on this Mac) and 'Logic?' (can this chain travel into the one fixed Assistant Scene Rack, or does it need host plug-ins). A scene is a document; third-party plug-ins are referenced by component id and never copied; separation results stay at family level, with the user choosing a candidate before it becomes a source. The engineering discipline is the interesting part. A module cannot declare its own execution lane — lane, ports, ranges and latency come from a registry that mints every manifest through a private initialiser. Readiness has no field in the document at all; it is computed from registration or from probing the live plug-in. Control values reach audio only as a flat, clamped snapshot on a 25 Hz tick, so a slider, an armed eye-tracking route, a MIDI controller and an approved AI proposal all arrive by the same door or not at all. Arming is not stored in the file, so opening a scene can never arm a sensor. The honest state is that this is a thoroughly source-complete and build-verified subsystem that has never been exercised as live audio. It compiles into the app, its document and canvas layers pass two headless suites that bite under mutation, the rack's masker kernel passes a ground-truth self-check at eleven block sizes, and the Voice Memo handoff has passing unit tests — but nobody has yet sat down, pressed play and listened to a scene end to end. The capability table distinguishes those levels rather than blurring them.

The 18 October 2026 media update adds actual native screenshots of the routing canvas and NoiseColour Character inspector. The graph and parameter controls are visible in the running app, with the engine idle. No microphone was started and no scene was played, rendered or exported during capture. These images establish the native interface, while the end-to-end audio and Logic export limitations below remain open.

Features & evidence

Workflow walkthrough

A guide to the native workflow. These steps describe the app; only the recordings and images embedded on this page are available as website demonstrations.

1. Open Audio Scene from the Panels menu, or the Audio Scene tile in the creative gallery. The palette on the left is grouped by execution lane, so you can see what KIND of work a module is before you pick it.
2. Add a source — an audio file, a Voice Memo, the microphone or system audio — then drag NoiseColour Character and an Output Bus onto the canvas and wire source into character into bus.
3. Read the verdict strip. It states separately whether the scene plays on this machine and whether it could travel to Logic, naming any module that would block either.
4. Pin a macro to the character stage's parameters, arm it and move it. Values reach audio only through the clamped snapshot, and the inspector shows the route and its health.
5. Search the plug-in browser for an installed Audio Unit by name and place it; it is probed as it lands, which reads its real parameter list and turns it from 'found' into usable.
6. Export. The scene writes a document, a restricted .assistantRackPreset, rendered assets and a Logic handoff note — or refuses in words, naming the stage the rack cannot take.