Pipelines & approaches for AI song dubbing. Pick a version.
Canonical orchestrator — the modular DAG (stems → transcription → melody → alignment → translation → lyric-gate → synthesis → merge) on Modal, with per-step benches. Contract-typed end to end.
Vocoder synthesis — PyWorld F0 transfer + Seed-VC timbre conversion (no language restriction). Reuses stems/transcription/translation/merge; own orchestrator + benches. Contract-typed end to end.
lab.moozz.com · moozz Modal workspace · moozz-lab bucket