using faster-whisper vs whisperx / model_v3?

[Andre3424]

hi!
noScribe is a really nice tool you made! We are using whisperX and I just compared now the transcription, while there are some differences, they seem really small. I was wondering: Why did you choose to use faster-whisper instead whisperx? One difference mentioned in discussions seems to be a more accurate timestamp with whisperX, which would be beneficial for your editor?

And out of curiosity: Did you compare transcriptions using model-v2 vs model-v3 of whisper? We did it, and the model-v3 quality was always without exception worse, than model-v2. I dont understand why, so I just wondered whether you could observe the same?