Header robot: AI Auto VJ — off / AI Auto VJ — running.

AI Auto VJ runs 100% on-device using Edge AI with zero network latency, zero credit cost, and complete offline privacy.

A rule engine reacts every bar to the audio energy state — ambient / building / peaking / breakdown. Building pushes the crossfader up, peaking cycles clips on every channel, breakdown blacks out on the drop or ducks opacity, ambient drifts back to center and cycles slowly. Actions are bar-quantized, max 3 per bar.
The on-device AI director is consulted on a configurable interval (4–64 bars, default 16) and sets pace and transition style — cut, crossfade, or hold. It synthesizes acoustic frequency bands, formant vocal detection, visual luminance and motion flux, and empirical learned channel slot roles.

The HUD displays live telemetry: {N}-bar phrase clock, energy section badge, multi-band spectrum, audio AI perception, visual telemetry, and director cues.

Energy comes from live audio capture — start capture first or AI Auto VJ has nothing to react to. While AI Auto VJ is running it learns from whatever is on screen and the live audio, then uses that to pick clips and FX. When AI Auto VJ is active, manual panel inputs are safely locked to prevent live accidental overrides.

The Master Live Composite header has a preview target selector — switch between Master, any configured Screen, and 3D Stage (drag to orbit) while the AI drives the show. Surface Control and Bus O overlays follow Screen and 3D Stage the same as they do on the projectors.