Critical Bug..

[legion7298]

Hello, First of all, I would like to express my deep respect for your work. Thank you so much for sharing such an amazing project. I have installed whisper-WEBUI via Pinokio, and I have successfully confirmed its operation through the web interface.

Let me first describe my working environment:

1. Operating System: Windows 10
2. GPU: Nvidia RTX 4070
3. CPU: Intel i9-9990K
4. RAM: DDR4 3200 32GB x 4EA = 128GB

Bug Report: As shown in the attached video, I selected the Whisper Large v3 model and used reasonable (default) settings to generate subtitles. However, I encountered a severe hallucination issue in the subtitles after a certain point in the timeline. For reference, when using Whisper CLI with VAD enabled, I did not experience such a critical issue in the output. I hope this report can help you improve this wonderful program and make it even more perfect.
Best regards.

[Audio]
http://encoding.legion-ms.com/sh25020401_track2.aac

[Bug Video]
https://github.com/user-attachments/assets/d3377ccd-6b87-4509-826a-e5b4efad9394

[jhj0517]

Hi, Thanks.
Can you link me the Whisper CLI project?
I think a different VAD implementation caused this, I should look into it.

[legion7298]

I leave a link that you can refer to.
This is the cli link I used. https://github.com/absadiki/subsai
If only this hallucination problem is improved,
I think it will be the most complete project of the web-ui currently released.
Thank you very much for revealing this wonderful technology.