search

Const-me / Whisper

High-performance GPGPU inference of OpenAI's Whisper automatic speech recognition (ASR) model
Mozilla Public License 2.0
8.2k stars 702 forks source link

Support for whisper-large-v2? #166

Open CharreyW opened 1 year ago

CharreyW commented 1 year ago

Hello there. Thank you for making this awesome app.

I want to ask about using whisper-large-v2 model. I got an error when i tried to use this model: https://huggingface.co/openai/whisper-large-v2

image

I tried using the recommended ggml-medium.bin (1.42GB) and it worked just fine. Thank you for your time and effort.

emcodem commented 1 year ago

There is no way to use original models from whisper or faster-whisper directly, instead you have to convert them to ggml format or just use the existing ggml models. Note that largev2 is nowdays often referred to as just "large". Described here: https://github.com/ggerganov/whisper.cpp

Look in "ggml format" section to find details about downloading existing ggml format models and for infos about how to convert models to ggml.

CharreyW commented 1 year ago

I see. Thank you very much for your answer.