Would really love if you could do Direct Whisper STT

[architect0fgod]

I was looking for a mobile keyboard that can use my whisper server for STT and deskdrop was just perfect. The ai features are really good but most of the time i just want Speech to Text directly with no changes but there is no option for it or to disable the llm model in ai voice. Basically a feature that allows to directly send whisper voice input directly instead of ai to modify.

[SvReenen]

Thank you for your message! Currently busy on another project but in a few more weeks im going to release another update for Deskdrop. I'll have a look into your suggestion and will report back here!

[architect0fgod]

Hello There, Appreciate your reply. I also wanted to bring something to your attention i was looking at some open weight models for STT and found something called moonshine and realized that this is like really fast compared to whisper and the moonshine streaming medium which is 274M parameter only takes few seconds to convert jfk speech of 11 seconds in just 3.6 seconds while something like whisper distill large v3 takes around minimum 30 secs and has lower WER