Transcription of calls from trunk-recorder using OpenAI Whisper.
- transcribe.sh runs from trunk-recorder which makes a POST request to the API, passing along the call WAV and JSON
- API creates a new task to transcribe the call and adds it to a queue (in RabbitMQ)
- Worker(s) (running on a machine with a GPU) picks up the task from the queue and executes it, transcribing audio
- As part of the task, worker makes an API call to Meilisearch to add the transcribed call to the search index
- If notifications are configured, as part of the task the worker will send appropriate notifications
If you're using OpenAI Whisper, you can use a local GPU to accelerate computations.
5054 links, including 361 private