How freyzamarshall02/voice-changer works
This Colab-focused realtime voice changer accepts an RVC model URL, loads the model on the server, and converts incoming audio for a browser client. The main runtime path spans browser audio capture and worklet processing, server request handling, model-slot setup, and RVC inference; the README specifically highlights URL downloads and RVC-only support. The source index and excerpts establish several internal calls, but do not expose the complete request/audio transport wiring, so those links are omitted rather than inferred.
Main entry points: PipelineGenerator.py Pipeline creation · VoiceChangerManager.py Voice changer manager · Pipeline.py RVC conversion · VoiceChangerClient.ts Voice changer client
Components 15
Browser audio
- Voice changer client
client/lib/src/VoiceChangerClient.ts - Audio worklet node
client/lib/src/client/VoiceChangerWorkletNode.ts
Server runtime
- Server startup
server/Okada.py - REST API
server/restapi - Socket.IO service
server/sio - Voice changer manager
server/voice_changer/VoiceChangerManager.py - URL model download
server/downloader/UrlModelDownloader.py
RVC model pipeline
- Model slots
server/voice_changer/ModelSlotManager.py - RVC slot generation
server/voice_changer/RVC/RVCModelSlotGenerator.py - Pipeline creation
server/voice_changer/RVC/pipeline/PipelineGenerator.py - RVC conversion
server/voice_changer/RVC/pipeline/Pipeline.py - Inference selection
server/voice_changer/RVC/inferencer/InferencerManager.py - Feature extraction
server/voice_changer/RVC/embedder/EmbedderManager.py - Pitch extraction
server/voice_changer/RVC/pitchExtractor/PitchExtractorManager.py
Other components
- User
Connections 18
- User → Voice changer client: uses
README.md - Voice changer client → Audio worklet node: creates
client/lib/src/VoiceChangerClient.ts - Voice changer client → Server startup: sends audio
- Server startup → REST API: starts
server/Okada.py - Server startup → Socket.IO service: starts
server/Okada.py - REST API → Voice changer manager: delegates to
server/restapi/MMVC_Rest_Fileuploader.py - Voice changer manager → URL model download: requests download
server/voice_changer/VoiceChangerManager.py - URL model download → Model slots: updates slots
server/downloader/UrlModelDownloader.py - URL model download → RVC slot generation: generates metadata
server/downloader/UrlModelDownloader.py - Voice changer manager → Model slots: manages
server/voice_changer/VoiceChangerManager.py - Voice changer manager → Pipeline creation: creates pipeline
- Pipeline creation → Inference selection: selects model
server/voice_changer/RVC/pipeline/PipelineGenerator.py - Pipeline creation → Feature extraction: selects extractor
server/voice_changer/RVC/pipeline/PipelineGenerator.py - Pipeline creation → Pitch extraction: selects extractor
server/voice_changer/RVC/pipeline/PipelineGenerator.py - Pipeline creation → RVC conversion: constructs
server/voice_changer/RVC/pipeline/PipelineGenerator.py - RVC conversion → Inference selection: runs inference
server/voice_changer/RVC/pipeline/Pipeline.py - RVC conversion → Feature extraction: extracts features
server/voice_changer/RVC/pipeline/Pipeline.py - RVC conversion → Pitch extraction: estimates pitch