Transcription endpoint needs to be reworked to meet the following requirements:
One single global transcription process, from a single source of audio
Endpoints for starting/stopping the global transcription process
A subscription/streaming endpoint that adds a client as a listener to the transcriptions
The subscription method can be as simple as keeping a list of clients, each with a unique web socket, that transcriptions are pushed to whenever generated.
Looks like I'll spend all break on this. I'll have to manually encode/transfer/decode/transcribe audio, which will involve a lot of audio processing that I know nothing about.
Transcription endpoint needs to be reworked to meet the following requirements:
The subscription method can be as simple as keeping a list of clients, each with a unique web socket, that transcriptions are pushed to whenever generated.