wispher.cpp with large-v3 model only use CPU，do not use GPU

[Aridea2021]

Dear developer, Many thanks for offering such a useful software, I try it from my friends sharing. now it seems that it can work with only CPU when I choose whishper.cpp large-v3 model. But I'd like to know how to use the GPU to accelerate the transcribe progress.

best regards.

from Dong

[raivisdejus]

If you have a GPU with enough VRAM there is little speed difference among Whisper.cpp and Faster Whisper. So the easiest way to get faster transcription with GPU is to use some other Whisper type.

[Aridea2021]

If you have a GPU with enough VRAM there is little speed difference among Whisper.cpp and Faster Whisper. So the easiest way to get faster transcription with GPU is to use some other Whisper type.

Hi raivisdejus, thank you for the kindness help, but I did not get you. I do have a modified 2080ti with 22g VRAM. It can work well in WhisperDesktop.exe with Whisper.cpp large v3 model. Can you tell me which whisper type should I choose so that I can use GPU to boost the progress. Thanks again~

[raivisdejus]

Use Faster Whisper.

But you also need CUDA installed https://developer.nvidia.com/cuda-downloads

[Aridea2021]

Use Faster Whisper.

But you also need CUDA installed https://developer.nvidia.com/cuda-downloads

so nice, thank you, raivisdejus