Spaces:

aadnk

/

whisper-webui

like 332

Running

Provide more info, what you are trying to do, what files are you using, where are you running it. Otherwise, it's not possible to determine where's the problem.

Provide more info, what you are trying to do, what files are you using, where are you running it. Otherwise, it's not possible to determine where's the problem.

I'm using the huggingface space on Opera browser under macos 10.13.6, on several 2-3 minute audio files in English.

@aadnk
I just tried on my Firefox browser, to upload a 13MB file, 0:57 long. In huggingface page it gives error after a little bit (i used medium model, VAD:none)

I tried on Google Colab - it works. So it's a problem in the huggingface page only.

I suggest running whisper in google colab, it's much faster anyway.

I personally run faster-whisper like this in google colab:

!git clone https://huggingface.co/spaces/aadnk/faster-whisper-webui
!cd faster-whisper-webui/ && git pull origin
!cd faster-whisper-webui/ && pip install -r requirements-fasterWhisper.txt
!cd faster-whisper-webui/ && python app.py --whisper_implementation faster-whisper --input_audio_max_duration -1 --share True --auto_parallel True

Hm, I just ran a 6:45 audio file (9.36 MB) onwhisper-webui on my HuggingFace page with no issue (using Windows 10/FireFox).

I was also able to run inference on the same file using MacOS 10.15.7 (I don't have an older version) with Opera.

But yeah, I agree with @Robis - you should try Google Colab or the Docker version.

I just tried again, and it works now. Maybe it was some non-space related issue with page.

Works now, thank you all