A sample program that combines sililero-VAD and faster-whisper for speech recognition. I use a browser because it was very cumbersome to handle microphone input directly from Python.
GPU and Cuda environment is required, please rewrite nvidia library dependencies in requirements.txt according to the CUDA version.
pip install -r requirements.txt
uvicorn app:main
Open a web page in your browser and speak into the microphone. The default microphone is used, so check your settings. Also, I don't know why, but the microphone rarely stops working in the browser. If this happens, please restart your browser.