How dasilva333/airi-audio-server works
AIRI Audio Server is a Node.js API for local C++ audio inference. Clients send speech, voice, music, or sound-effect requests to the HTTP server; routes dispatch work to the relevant engines and return audio, metadata, or streaming responses. Voice registration adds normalization and optional transcription, while a shared GPU queue coordinates GPU tasks. Model setup/download is a separate provisioning capability. Relationships for sampled code are cited; broader capabilities and queue behavior are included where documented in the README.
Main entry points: server.js HTTP Server · routes.js API Routes · voices.js Voice Catalog · gpu.js GPU Binary Resolver
Components 16
API and Coordination
- HTTP Server
src/server.js - API Routes
src/routes.js - GPU Queue
src/queue.js - Text Processing
src/text.js
Speech and Voices
- Speech Engine
src/engine.js - Voice Catalog
src/voices.js - Speech Recognition
src/stt.js - Audio Normalization
src/ffmpeg.js
Audio Generation
- Music Engine
src/music.js - Sound Effects
src/sfx.js - Voice Design
src/voice_design.js
Inference Runtime
- GPU Binary Resolver
src/gpu.js - audio.cpp Inference
Model Provisioning
- Model Setup
setup.js
Other components
- API Client
- Hugging Face
Connections 27
- API Client → HTTP Server: sends requests
README.md - HTTP Server → API Routes: mounts routes
src/server.js - HTTP Server → GPU Queue: creates queue
src/server.js - HTTP Server → Text Processing: creates processor
src/server.js - HTTP Server → Speech Engine: creates engine
src/server.js - HTTP Server → Voice Catalog: creates manager
src/server.js - HTTP Server → Music Engine: creates engine
src/server.js - HTTP Server → Voice Design: creates designer
src/server.js - HTTP Server → Sound Effects: creates engine
src/server.js - API Routes → Speech Engine: dispatches speech
README.md - API Routes → Voice Catalog: registers voices
src/routes.js - API Routes → Music Engine: requests music
src/routes.js - API Routes → Audio Normalization: converts uploads
src/routes.js - API Routes → Speech Recognition: transcribes audio
src/routes.js - API Routes → GPU Queue: queues GPU tasks
README.md - Speech Engine → GPU Binary Resolver: resolves binary
src/engine.js - Speech Engine → audio.cpp Inference: streams inference
src/engine.js - Music Engine → GPU Binary Resolver: resolves binary
src/music.js - Music Engine → audio.cpp Inference: runs inference
src/music.js - Sound Effects → GPU Binary Resolver: resolves binary
src/sfx.js - Voice Design → GPU Binary Resolver: resolves binary
src/voice_design.js - Voice Design → Voice Catalog: ingests generated voice
src/voice_design.js - Voice Catalog → Audio Normalization: normalizes clips
src/voices.js - Voice Catalog → Speech Recognition: transcribes references
src/voices.js - Speech Recognition → Audio Normalization: normalizes input
src/stt.js - Speech Recognition → GPU Binary Resolver: resolves binary
src/stt.js - Model Setup → Hugging Face: downloads weights
README.md