Cost-optimized Speech-to-Text WebSocket server with GPU acceleration.
wss://pgits-stt-gpu-service-v3.hf.space/ws (external)ws://localhost:7860/ws (local development)
// Start streaming
{"type": "start", "config": {"enable_timestamps": true}}
// Send audio (base64 encoded)
{"type": "audio", "data": "...", "sample_rate": 16000, "channels": 1}
// Stop streaming
{"type": "stop"}