VoiceForge uses the main server plus optional model services. Audio processing runs inside the main server:
- Main API, Audio Services, RVC:
app/install/install_main.bat,Voice_Forge.bat - ASR (Whisper + GLM-ASR + Parakeet):
app/install/install_asr.bat,app/launch/launch_asr_server.bat
Background stream mixes are cache artifacts and are stored outside output/ by default in the system temp directory. Override with VOICEFORGE_BG_STREAM_CACHE_DIR if needed.
Text chunking is owned by Extension-Voiceforge. The API expects one already-chunked TTS segment per request and rejects oversized segments instead of splitting them server-side.