I’ve been using the Whisper API for some time, and I’ve noticed that it’s been acting “lazy.” It’s skipping important parts of the transcription, which didn’t happen before (I tested it on a model installed on my local machine, and the transcription is perfect, with 100% success in the transcription).
Furthermore, it seems to be random because if I try to transcribe the same audio file again, sometimes it transcribes the part it couldn’t transcribe in the previous attempt.
Is anyone else experiencing this as well? Could there have been a model update causing the transcription to worsen?
I transcribe phone calls, so I believe it wouldn’t fall under copyright issues.
Just as a reference, I use the Whisper API in the Portuguese language.
1 Like
Hey! I had the same issue and fixed it by adding 5 seconds of blank audio at the end of the audio. Somehow this works, only thing is it adds some weird credits at the end, like “Subtitles by …”.
When I ran whisper locally I “fixed” it by switching to the medium model.