Whisper model in Sentis with audio samples over 30s?

AI & Machine Learning AI Inference Engine
, ,
1

Hi, I’m using whisper with sentis and I’ve come across the need to use audio samples longer than 30 seconds.

I understand that it’s possible to use chunking algorithms but I’m not entirely clear on how this is possible in Unity.

Has anyone been able to pass audio longer than 30 seconds?

Regarts!

openai/whisper-tiny · Hugging Face (long form transcript)

2

That’s a specific of whisper tiny model I’m afraid.
If they removed that requirement from their architecture, re-export the model and have a go at it!