Probability or score coming from faster-whisper and not alignment model

[arbianqx]

Hello everyone,

Hope you're doing well.
Is there any chance that we can get probability or score of words coming from faster-whisper itself and not from alignment model.
The reason behind this, is that score coming from alignment model is not reliable as the faster-whisper, which looks more correct.

Any approach, on how we might achieve this?

Thanks a lot for your time