How msrbuilds/voice-studio works
Voice Studio is a local web application for speech generation and audio-to-audio workflows. Users choose a project mode and voice, then synthesize speech, transcribe audio, translate text, or dub a clip; generated audio and transcripts can be edited, cached, and exported. The backend coordinates selectable TTS engines, ASR, translation, voice management, and model lifecycle operations. The README documents the major modes and capabilities; some subsystem connections are therefore documented rather than verified from sampled implementation.
Main entry points: app.py Application and routing · synthesize.py Synthesis orchestration · dub.py Dub orchestration · transcribe.py Transcription service
Components 25
Studio UI
- Project workspace
frontend/src/App.tsx - Mode editors
frontend/src/components - Engine selector
frontend/src/components/EngineSelector.tsx - API client
frontend/src/lib/api.ts
HTTP API
- Application and routing
backend/app.py - Synthesis and streaming routes
backend/api - ASR and transcription routes
backend/api/asr.py - Dubbing route
backend/api/dub.py - Translation route
backend/api/translate.py - Engine management routes
backend/api/engines.py - Voice routes
backend/api/voices.py
Speech workflows
- Transcription service
backend/services/transcribe.py - Whisper ASR
backend/core/asr/whisper_engine.py - Dub orchestration
backend/services/dub.py - Offline translation
backend/services/translate.py - Synthesis orchestration
backend/services/synthesize.py
Models and voices
- Engine manager
backend/core/engine_manager.py - TTS engine implementations
backend/core/engines - Voice registry
backend/services/voices.py - Model lifecycle
backend/services
Caches and services
- Synthesis cache
backend/services/synth_cache.py - Workflow caches
backend/services - GPU execution gate
backend/core/gpu_gate.py
Other components
- Local user
- Local audio files
Connections 31
- Local user → Project workspace: uses
- Project workspace → Mode editors: renders
frontend/src/App.tsx - Project workspace → API client: requests
- Engine selector → API client: requests
- API client → Application and routing: calls API
- Application and routing → Synthesis and streaming routes: registers routes
backend/app.py - Application and routing → ASR and transcription routes: registers routes
backend/app.py - Application and routing → Dubbing route: registers routes
backend/app.py - Application and routing → Translation route: registers routes
backend/app.py - Application and routing → Engine management routes: registers routes
backend/app.py - Application and routing → Voice routes: registers routes
backend/app.py - Synthesis and streaming routes → Synthesis orchestration: delegates to
backend/api/deps.py - ASR and transcription routes → Transcription service: delegates to
backend/api/asr.py - Transcription service → Whisper ASR: transcribes with
- Dubbing route → Dub orchestration: delegates to
backend/api/dub.py - Dub orchestration → Transcription service: transcribes via
- Dub orchestration → Offline translation: translates via
- Dub orchestration → Synthesis orchestration: synthesizes via
- Translation route → Offline translation: delegates to
- Engine management routes → Engine manager: manages engines
backend/api/engines.py - Engine management routes → Model lifecycle: starts lifecycle tasks
backend/api/engines.py - Voice routes → Voice registry: manages voices
- Synthesis orchestration → Engine manager: selects engine
backend/services/synthesize.py - Engine manager → TTS engine implementations: constructs engines
backend/core/engine_manager.py - Synthesis orchestration → Voice registry: resolves voice
backend/services/synthesize.py - Synthesis orchestration → Synthesis cache: reads and writes
backend/services/synthesize.py - Synthesis orchestration → GPU execution gate: runs inference through
backend/services/synthesize.py - Transcription service → GPU execution gate: serializes inference
- Dub orchestration → Workflow caches: uses workflow caches
- Local audio files → ASR and transcription routes: uploads audio
- Local audio files → Dubbing route: uploads audio