snakers4/silero-vad ? reverse-engineered prompt

Build me a small Python package that can tell when a person is speaking in an audio file, and give me the start and end times of each speech segment. I want it to be easy to use from a simple script, but also work in a notebook, with a few examples showing how to load audio, run detection, and print the timestamps.

Please make it fast and lightweight, with support for both normal Python use and a model file that can run through ONNX too. It should handle common audio formats, work well on short clips and longer recordings, and be useful for things like call audio, voice bots, and cleaning up speech datasets. If you need to, look up the current docs for the best way to load and run the model. Also include basic tests so I can tell it’s working.