Ichigo
StaleDescription
Open-source local realtime voice AI system supporting fully offline real-time voice conversations, suitable for building private voice assistants and voice interaction applications.
Key Features
- Fully offline real-time voice AI system with ASR, LLM, and TTS capabilities
- Compact 22M parameter speech tokenizer for multilingual ASR (English + Vietnamese)
- Early fusion technique inspired by Meta's Chameleon for speech-language model integration
- Batch processing support for single files and folder-based transcription
- FastAPI service for frontend integration with OpenAI-compatible API endpoints
- Modular architecture enabling cross-task training between ASR and TTS components
Use Cases
💡 Building private voice assistants with fully local inference
💡 Multilingual speech recognition for offline deployment scenarios
💡 Research into speech-language model unification and early fusion techniques
💡 Creating voice interaction applications with ASR and speech language models
Categories
Quick Start
Install via 'pip install ichigo'. For ASR, use 'from ichigo.asr import transcribe; results = transcribe("path/to/file")'. For batch processing, pass a folder path. To start the API server, run 'uvicorn asr:app --host 0.0.0.0 --port 8000' or use Docker. Access the API docs at localhost:8000/docs.