Hervé BREDIN comments

Results 270 comments of


                                            Hervé BREDIN

In pyannote voice activity detection task, fluctuating huge memory?

I haven't really thought about memory usage up to now. If you happen to profile the code, I'd appreciate if you could let me know about your findings.

inference call not accepting clean wav audio

Would you mind sharing a link to a Google Colab containing a self-contained minimum reproducible example that I can run to reproduce the error?

VAD Significant inconsistency in the results when using CPU and GPU

Thanks for reporting this. Would you mind running the following and sharing the visual output on both CPU and GPU? ```python segmentation = Inference(model, skip_aggregation=True) segmentation("v.wav") ```

VAD Significant inconsistency in the results when using CPU and GPU

Any update on this?

Speaker-diarization acceleration

You can indeed [increase the batch size](https://github.com/pyannote/pyannote-audio/blob/26cddd98f91a770f9c73f569f200674fda1ee705/pyannote/audio/pipelines/speaker_diarization.py#L76 ) Do not expect a huge speed up, though.

Diarization and Database recipes

Tests effectués sous Chrome (macOS) **Problème #1** [H] affiche l'aide. [H] cache l'aide. [H] affiche l'aide. [ESC] cache aussi l'aide mais "skip" le "chunk" courant et passe au suivant. Il...

Diarization and Database recipes

**Problème #9** Il semble que l'audio snippet (qui est joué quand on passe la souris sur les étiquettes en haut) contient des extraits audio qui ne sont pas dans l'ordre...

Limiting number of speakers?

This is not yet possible but is something I have (not that deep) in my TODO list.

Limiting number of speakers?

The pretrained pipeline in pyannote.audio 2.0 now supports this option. See documentation [here](https://huggingface.co/pyannote/speaker-diarization).

Interactive pyannote.audio.utils.preview

https://gist.github.com/scottire/654019e88e6225c15a68006ab4a3ba98 https://webaudioconf.com/_data/papers/pdf/2018/2018_19.pdf