alesaccoia/VoiceStreamAI

Employs Huggingface's pyannote VAD alongside faster-whisper for optimized speech detection and transcription, with factory/strategy pattern design enabling modular swapping of VAD and ASR components. Handles chunked audio streaming with dynamic silence-aware buffering to prevent mid-word cutoffs while minimizing latency, and supports optional SSL/TLS encryption for secure WebSocket connections. Includes Docker deployment with GPU acceleration support and multilingual transcription capabilities across both Python server and browser-based JavaScript client.

950 stars. No commits in the last 6 months.