Difference between result of python api and command line usage

[lionsheep0724]

I saw a difference between python and command line usage.(low-high level of python, command line are all different) My audio file is about 3.5MB.
I expect that their arguments are somewhat different. Any ideas here?

[lionsheep0724]

I found more strange thing!
whisper throws diffrent results every single run, with same .wav file.

[jzhengdj]

When doing the translate task (Japanese to English), sometimes one sentence will be repeated multiple times, causing the generated subtitle to be wrong for 10-30 seconds. Wondering how to solve this issue.

[lionsheep0724]

Many thanks to your reply!
BTW, I'm doing transcribe task, in vscode notebook. Even running for loop as below shows diffrent result as loop goes.
for _ in range(5): result = model.transcribe(audio_path) print(result["text"])

[glangford]

I saw a difference between python and command line usage.(low-high level of python, command line are all different) My audio file is about 3.5MB. I expect that their arguments are somewhat different. Any ideas here?

The difference between command line and python execution is explained in this discussion thread

• Much slower via command line tool than in Python? #177

I found more strange thing! whisper throws different results every single run, with same .wav file.

On the issue of different results across runs, see

• Different transcribe results using same model to transcribe the same audio #81