[feature request] audio tagging with Whipser-AT

[phineas-pta]

so somebody manage to integrate audio tagging with whisper:

• announcement: 🚀 Introducing Whipser-AT, a new joint audio tagging and speech recognition model. openai/whisper#1504
• source code: https://github.com/YuanGongND/whisper-at

it seems they dont modify original whisper architecture (load directly openai checkpoints), the audio tagging feature is somehow hooked into it and load their pretrained checkpoint

it's unlikely the author could implement faster-whisper: YuanGongND/whisper-at#1

[Ar770]

Any plans to add this feature?

[kuroderuta]

Yeah, this would be really useful.