Scope#
V2 uses @v2/isis-commentary-lipsync-pipeline to turn AI commentary audio and
transcripts into commentator face curves. This closes the prior Audio2Face gap
by wiring @isis/visual-dubbing as the V2 commentary lipsync provider.
Contract#
The pipeline runs:
- commentary audio ingest
- phoneme extraction
- visual-dubbing lipsync
- lip-sync quality validation
- Unreal curve export
Outputs target:
- disk root:
V2/ue/Content/V2/Audio/Commentary/Lipsync - Unreal package root:
/Game/V2/Audio/Commentary/Lipsync
Rendered payloads are tagged v2.commentary.lipsync and include commentary line
id, speaker id, locale, content root, Unreal package root, package name, and the
maximum allowed lip-sync drift.
Rollback Boundary#
Commentary lipsync is audio/visual only. Plans carry
deterministicGameplay: false and
rollbackPolicy: "off-rollback-audio-visual-only". V2Audio may play or discard
rendered curves, but those outputs never feed gameplay simulation or rollback
state.