Limit to 30s

[thewh1teagle]

I'm using pyannote-onnx in conjunction with whisper.cpp, and I'm encountering an issue where whisper.cpp expects audio clips to be no longer than 30 seconds. However, pyannote sometimes detects speech longer than 30 seconds.

I tried splitting the speech chunks that exceed 30 seconds, but this sometimes results in cuts right in the middle of spoken words. How can I handle this more effectively to ensure that each audio segment is at most 30 seconds without splitting words awkwardly?

[pengzhendong]

Hello, perhaps you could consider employing some rules such as silero-vad.

https://github.com/snakers4/silero-vad/blob/1d48167271d1ba3517107527f09063fdf374707b/src/silero_vad/utils_vad.py#L218-L230