Real-time speech-to-text: architecture of a live transcription feature

Streaming audio from the browser over WebSockets, chunking for low latency, partial versus final transcripts, a post-recording accuracy pass, and the