This is a basic python program, which uses ffmpeg to split an audiostream into 5 second clips.
The clips are then run through whisper, via the python library.
Performance of this programm is as fast as the system can handle whisper.
FFmpeg,
Python3.10 (whisper only supports 3.10 for now),
whisper installation instructions can be found here: https://github.com/openai/whisper
If your intention is to run with cuda, ensure the correct cuda-included PyTorch version is installed.