M0 realtime feasibility lab

Musical audio path diagnostics

Permission-driven device capture, a stable Voice / Instrument / Mix track, direct-or-TURN P2P media, and privacy-scrubbed connection health.

Request access to begin.
1

Permission and devices

Device enumeration starts only after this user action. Labels remain in this page and are never included in logs or copied diagnostics.

Requested voice processing
Voice EC: unavailableVoice NS: unavailableVoice AGC: unavailableVoice sample rate: unavailableInstrument EC: unavailableInstrument NS: unavailableInstrument AGC: unavailableInstrument sample rate: unavailableAudioContext: unavailable Hz
2

Stable outbound audio

Track stable: no

Focus this panel and press V, I, or M. Switching changes gains; it never replaces the outbound track or creates an offer.

NAM: not prepared

Only the instrument enters NeuralAmpModelerCore. The voice branch is untouched. Failure code: none; loaded preset: none. These bundled examples are intended to exercise model parsing and inference; they are not production tone presets or an audible-quality rating.

Voice0%
Instrument0%

SDP offers emitted by this session: 0

3

1:1 P2P session

not-created

No temporary TURN credential loaded.

Local preview
Remote participant
4

Connection diagnostics

CodecUnavailable
Audio bitrateUnavailable
Packet lossUnavailable
JitterUnavailable
Round-trip timeUnavailable
ICE candidateUnavailable
Media pathUnknown
Secure contextUnknown
5

NAM/WASM feasibility boundary

Boundary fixture: non-NAM

This measures an AudioWorklet/WASM boundary with a self-authored gain fixture. It verifies main-thread-free processing, explicit bypass, and fail-open behavior, but it is not a Neural Amp Modeler engine or model. A separate 48 kHz prototype now has a pinned engine and two allowlisted test models; representative performance, parity, audible-quality, and redistribution evidence remain open.

Run from a secure browser context. Audio output is muted during the boundary benchmark.