add whisper turbo and real time with gradio 5.0

[AK391]

Hi would be great to add the new whisper model https://huggingface.co/openai/whisper-large-v3-turbo and also enabled real time streaming with gradio 5.0 see: https://huggingface.co/spaces/KingNish/Realtime-whisper-large-v3-turbo

[jhj0517]

Hi @AK391, Thanks for noticing me update!

Since faster-whisper does not officialy support turbo yet,
you can download deepdml/faster-whisper-large-v3-turbo-ct2 and place it in Whisper-WebUI\models\Whisper\faster-whisper and use it for now.

About realtime streaming, I'll work on it soon! I'm just leaving a link that I'll refer to later -

• https://www.gradio.app/guides/real-time-speech-recognition#3-create-a-streaming-asr-demo-with-transformers

[OneSteelTank]

Hey, I'm a bit confused.

Is this how I'm supposed to do it? I went to the website you linked and downloaded the model.bin, but when I look back at the WebUI nothing changes.

[jhj0517]

@OneSteelTank Same with other faster-whisper model structure, placing the model files (including other metadata files) in the model name subdirectory is what you should do:

Then the subdirectory name faster-whisper-large-v3-turbo-ct2 will appear in the models dropdown.

[OneSteelTank]

@jhj0517 ohhh that makes much more sense. thank you so much man, I see it now.

[wheezy1749]

For docker noobs like me that found this through google. Didn't show up until I replaced in docker-compose.yaml

      - /Whisper-WebUI/models

with

      - /path/to/Whisper-WebUI/models:/Whisper-WebUI/models

Not sure if it's right but its how I got it working with docker. Don't need to rebuild. Just restart the container