Add normalization to sox call.

[xfim]

I believe it can improve the capacity to recognize patterns, from my rudimentary tests. But I am curious about what you think.

Another, more complex option, would be to do it to each of the small chunks of audio, one by one. In that case, I have not been able to find a suitable function on pyAudioAnalysis. But I am sure there must be one way to do it.

Thank you very much for the tool.

[abhirooptalasila]

Hi
Sox has a default norm value of -3.0. I think the value you've provided in the CLI argument works better for your use case.
How big of a difference is visible in your generated subtitles?

[xfim]

Honestly, I have tried now and explicitly using -3.0 gives slightly (very slightly) better values than my proposal of -0.1. But I have to admit that I have used it / adapted it not in the context of a movie with arranged audio, but using more domestic / low quality videos with poorer audios.

Comparing using -0.1 and nothing gives better results to the default (nothing), but not so well as -3.0.

So I leave it to you what to do with it. I am far a way from being an expert.

[abhirooptalasila]

I think -3.0 works best for a variety of use-cases. Thank you!

[xfim]

Thank you.