[Feature] Add support for insanely-fast-whisper over faster-whisper

[Timtam]

Description of the feature you are requesting:

I'm by far no expert, but I noticed that insanely-fast-whisper is now out which further reduces time to transcribe audio: https://github.com/Vaibhavs10/insanely-fast-whisper

Describe the solution you'd like

Switch from faster-whisper to insanely-fast-whisper as a permanent replacement.

Describe alternatives you've considered

Optionally enable the user to decide wether they want to run faster-whisper or insanely-fast-whisper via an environment variable or something similar.

Additional context

None

[Arche151]

Just tried it out and it's reeeally good! Also, it uses the large-v3 model by default, instead of large-v2 like faster-whisper. Aand the developer said, he will add diarization this week!

Edit:

Insanely-fast-whisper was about 4x faster than faster-whisper in my test, but it also uses much more VRAM than faster-whisper - about 13GB, compared to faster-whisper's 4,7GB.

So yeah, being able to choose which whisper-implementation to use in an environment variable would be great!

[pluja]

I will look into this, it should not be that hard to add it as an alternative whisper backend with the current transcription-api implementation.

I would like to keep faster-whisper as a backend since it allows the use of CPU, and insanely-fast-whisper only allows GPU. So it should be a dropdown that allows the user to choose which backend they want to use.