Audio Master

Workflow

1. Read ../../references/AUDIO_MASTERING.md, the edit plan, voice/music/SFX cues, media registry, and approvals.
2. Write or update 05_post/audio_mix.json with exact timeline starts, source trims, durations, gain, pan, and fades.
3. Reject unresolved or missing required media before final rendering.
4. Use scripts/audio_master.py PROJECT --manifest 05_post/audio_mix.json for the actual master.
5. Verify the output exists and preserve the renderer report.
6. Run asset-approval to register the rendered master.
7. Run delivery-qc when the master has delivery constraints.

Weak-model rule

Do not estimate a rendered waveform path. Do not treat cue descriptions as audio. Do not stretch exact dialogue merely to fill a picture gap.

Done

The project has a reproducible mix manifest and, when rendering was requested and FFmpeg succeeded, an actual synchronized audio master file.