Allow usage of ffmpeg audio filters

[acastin]

Allow users to use ffmpeg audio filters to clean up or normalize audio data, with a link to the official documentation and some examples

[Barabazs]

My 2 cents: this should be part of the users workflow, outside of the context of WhisperX.

1. it opens the door to shell/binary exploits
2. additional dependency headaches since ffmpeg isn't shipped with WhisperX
3. shell errors/exceptions are hard to handle and will result in a subpar experience

[NielsMayer]

@acastin -- what kind of results and improvements are you seeing when employing, e.g., "EBU R 128" or "dynaudnorm" ? Is there a reason regular whisper doesn't employ this technique? Do any other variants?

ffmpeg_audio_filters: Optional[str]
        Apply ffmpeg audio filters (https://ffmpeg.org/ffmpeg-filters.html)
        "ebur128" to normalize loudness across the audio based on EBU R 128
        "dynaudnorm=p=0.5:s=5:g=15" to normalize volume dynamically over short windows

Thanks for pointing out this feature of ffmpeg... looks quite useful!

[clort81]

As a rule: Stick to Unix Philosophy for maintainability -- Keep scope, bindings, dependencies limited. Yes, people are forgetting this wisdom, which is why I write it again and again. 'App thinking' is a commerce model. Not applicable in FOSS tool as here.
This is what scripting is for. Tool pipelines in text, graphics, whatever.
Btw 'normalize dynamically' is called compression in music engineering.