[feature request] whisper

[0wwafa]

Since you are the only one (my hero!) who still supports CLBLAST, and since you already use whisper, can you make kobold.cpp act like whisper.cpp, so I can use it to transcribe some long italian movies and translate them to english?

[LostRuins]

It already exists! You just need to load the whisper model with --whispermodel. See https://github.com/LostRuins/koboldcpp/wiki#what-is-whisper

KoboldCpp can also transcribe wav audio files. using the OpenAI compatible endpoint. For example:

curl --request POST \
  --url http://localhost:5001/v1/audio/transcriptions \
  --header 'Content-Type: multipart/form-data' \
  --form file=@/path/to/file/audio.wav\

[0wwafa]

I see. it would be somewhat more useful to have it directly like:
./kobold.cpp -wm model.bin -ot time_offset -p "additional prompt" input.wav

or something like that.
but I'll try it out in server mode.

[LostRuins]

Hi, can you please try the latest version 1.80, I've added a feature to upload files to transcribe from the GUI