Reconstruct audio after VAD

[kojomensahonums]

I have just used the pyannote-segmentation model for Voice Activity Detection. I very much liked how I was able to evaluate the model's performance. Now, I am seeking a way to use the same model but then to reconstruct the speech segment outputs into an audio. This will be similar to using the VAD model to filter out non-speech portions from an audio. Any help on this is greatly appreciated!