x86Gr

> It is a known issue with near zero signals. This is not a bug, this is a feature (tm). > > The newer VAD tries to suppress spurious activations...

> > Since the noise is really near-zero, could this be prevented by a preprocessing filter cutting long parts below a set threshold, or -20dB below the average RMS? >...

Will this work with generic AMD gpus, ie newer integrated gpus?

@pemistahl I don't know if this is related, but.. we process both short and long texts, and with short texts we observed a weird behaviour with some language combinations. I...

Interesting. Can you run that again with _LanguageDetectorBuilder.from_all_spoken_languages()_ ?

For me it is improving the accuracy a bit, but i'm using short sentences.

FYI, when using "without_timestamps=True" on faster-whisper 1.0.3 I get a faster speed but with a lot of skipped sentences.

I agree, I also have worse performance, just not as much, however the overall WER for non english speech is going down. Go back to silero or at least let...

I will run the tests on our audio corporas, with different parameters, but it won't be quick

Can it be made compatible with numpy 2.x ?