A server to give easy access to the whisper model using browser APIs.
Currently under development is a feature to forward the recognized speech to a language model.
- Specify the model size you need, in model.py. Note that the large model needs about 12 GB VRAM
- Create a docker image with the dockerfile, then run it:
docker build -t whisper .
docker run --gpus all whisper:latest
The server will now listen on port 80.