Skip to content

Continuity and Audio

Ethanfel edited this page Sep 1, 2026 · 3 revisions

Continuity and audio

Use Generation Profile for new workflows. It keeps the two user decisions separate: what the picture carries into the next scene, and what happens to sound.

Choose scene continuity

Choice Result Start here?
Visual continuity Carries a clean visual guide from the accepted predecessor. Yes
Independent scenes Carries no predecessor picture. For intentional cuts
Hard picture + protected audio Places an exact picture/audio prefix in the target. Advanced
Hard picture + smooth audio release Protects picture while releasing generated audio gradually. Advanced

The choice applies to the boundary entering a scene. It does not change the completed predecessor.

Advanced Policy Override exposes Tone, Latent, Detail, Drift-Control, and Color-Stable Drift behavior. Keep those supplied settings unless you are running a controlled comparison.

Choose an audio profile

Audio profile Result
Generate audio H3 creates the final sound and carries it between scenes.
Generate fresh audio per scene H3 creates sound without carrying the preceding audio latent.
Lip-sync to source audio Exact Source Timeline windows drive generation and remain the final soundtrack.
Generate audio from source guide Source sound guides H3, which creates the final sound.
Use source soundtrack only Source sound is used at assembly without guiding generation.
No final audio The final MP4 is silent.

For an exact song or dialogue track, add Lip-Sync Options only when you need pre-roll/lookahead control or a vocal timing mask. The original Source track still supplies final audio.

Source Timeline

Register a long source track once:

video/audio → Source Timeline → Loop Start
                              → Preflight

Current Shot selects the exact scene-local window. Project Asset Carousel can also store one asset as the project's Source track and pass it through Plan.

Trimming and output

Loop Trim removes repeated incoming picture context and conforms generated audio to delivered video timing. Connect its media outputs to Segment + Checkpoint.

Every connected generated soundtrack is retained as a WAV sidecar even when the final MP4 uses Source audio. Assemble normally follows the audio choice saved in Plan; its explicit source/generated/none settings are overrides.

Legacy policies

Manual Chain Policy (Legacy) remains available for existing graphs and unusual combinations. Legacy 0.4 Policy Adapter is for actual 0.4 imports. Prefer Generation Profile for new graphs.

The repository's detailed audio guide covers per-scene overrides, AV timing, lip-sync masks, and experimental modes.

Clone this wiki locally