search

resemble-ai / Resemblyzer

A python package to analyze and compare voices with deep learning
Apache License 2.0
2.66k stars 419 forks source link

trim out problem in diarization #63

Open HedgehogTW opened 2 years ago

HedgehogTW commented 2 years ago

The preprocess_wav function will trim out the silences in the audio file, and return a new wav, but it did not return information about which segments were cut. If we are dealing with a video, the audio timeline is different from the video frame timeline. Therefore, we cannot process the video frame according to the diarization. Can the preprocess_wav function also return some information, and then we can restore it to the original timeline?

teoh79 commented 2 years ago

Hello I face the same problem as I need to have the timestamps of the speakers from the original wav file: have you progressed ln this issue? Thank you