Can't transcribe a longer interview :(

[Pompey21]

Hi guys, I am trying to use Whisper for transcribing an interview (circa 30min long) but I only get the first 10 seconds of it - not even the first 30 as stated in the README file. I would much appreciate your assistance on figuring what I am doing wrong. Here is my 2-line code:

model = whisper.load_model('base')
result = model.transcribe('/Interview-Claude-Nobs.wav')

Much appreciate your help and assistance!

[jongwook]

Does the file have a long silence in the beginning? Whisper often struggles with silent parts, and trimming those might help.

Otherwise, can you try in the command line, or using different models, like base.en or small?

[marcpope]

I trandscribe 2 hour podcasts without issue. Using a GPU

[Pompey21]

Yes there might be a short (circa 2-3s) break. So you recommend I trim these?
Also, I cannot use GPU sadly...

[elementalcollision]

I've processed 2.5hrs of audio (almost 3 at this point) in 2 WAV files using --medium (it was a mistake) on an M1 MAX and while slow, it's chugging through. My issue seems to be extremely poor CPU utilization.

[jongwook]

2-3s breaks are usually fine, but if the issue persists even after trimming or even with other models, please feel free to upload/email the audio file if you can. Thanks.

[FurkanGozukara]

I have a video that explains how to use whisper it may help. i install on windows from scratch and run it

https://www.youtube.com/watch?v=msj3wuYf3d8