Performant & scalable speech-to-text and text-to-speech layers for any voice-conversational LLMs.
To run the WhisperLiveServer, cd into WhisperLiveServer and run:
docker build -f Dockerfile.gpu -t whisperlive-gpu .
docker run -it --gpus all -p 9090:9090 --rm whisperlive-gpu
docker build -f Dockerfile.tensorrt -t whisperlive-tensorrt .
docker run -it --gpus all -p 9090:9090 --rm whisperlive-tensorrt
docker build -f Dockerfile.cpu -t whisperlive-cpu .
docker run -it -p 9090:9090 --rm whisperlive-cpu
To run an example app that uses the WhisperLiveServer to transcribe, cd into app and run app.py.
/tools/ contains Go scripts that can be used to record and re-play a Twilio call's websocket stream.