It uses flask to wrap openai whisper into a http web api server
just clone the repo and
conda create -n whisper python=3.10
conda activate whisper
pip install -r requirements.txt
if your torch doesn't support cuda, just go to https://pytorch.org/get-started/locally/ and find the one with your system. for example, if you are running this on ubuntu, use this:
pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118
run it with the following command in windows
start.bat
you can also run this in linux, and it's the same
flask run --host=0.0.0.0 --port=5500
this is only for the purpose of testing, if you need to run this in a production environment, you should put this behind a proper http server like nginx and create a reverse-proxy. it's extremely easy to be hacked if you put this on a public ip. check this https://flask.palletsprojects.com/en/2.3.x/deploying/