Whisper Transcriber
Web client for the transcription gateway: stream over Socket.IO, or upload a file over REST.
Disconnected
Audio is resampled to 16 kHz mono Int16 and sent as transcribe_audio.
The mic only works on localhost or HTTPS.
Works like client.py: the browser decodes the file, cuts it into chunks and emits them one by one.
Sends POST /transcribe-file (wav, mp3, m4a, webm, ogg, up to 30 MB). The first request can take a minute while a GPU worker starts.