Audible demo delivery

The first audible slice now includes generated music, reviewed import, two to four actor rotation, model-independent observation contracts, Rig model calls, local-planner fallback, frame scheduling, actual stereo mixing, durable journals, WAV download and replay verification.

The default demonstration uses a transparent autonomous local planner. A model-backed run requires a configured backend and is reported separately. The full acceptance gate remains a 20–30 minute set using real reviewed tracks and independent model decisions, evaluated by listening as well as metrics.

Next work: validate the selected model with real music; replace provisional analysis with an evaluated Essentia/ONNX or alternative analyzer; improve harmonic/phrase/vocal transition scoring; then add realtime output. Hardware, club library export, effects and desktop packaging remain post-MVP.

Use the current implementation as the authoritative feature inventory.