Enable prompt broadcasting / specification
Opened this issue · 2 comments
Blair-Johnson commented
Currently prompt specification of the form model.transcribe([audio1, audio2], prompt='prompt')
is untested. Need to investigate and enable the following usage: model.transcribe([audio1, audio2], prompt=['prompt1', 'prompt2'])
, and model.transcribe([audio1, audio2], prompt='prompt')
(where 'prompt'
is used for all input files).
TODO
- Investigate current behavior when prompts are specified by the user
- Enable per-file prompt specification
- Enable broadcast prompt specification
shaynemei commented
@Blair-Johnson Hi, this would be a great feature. Is this being worked on?
Blair-Johnson commented
@Blair-Johnson Hi, this would be a great feature. Is this being worked on?
Thanks for your interest! I currently don't have a timeline for implementing prompting, but it is on the to-do list for when I have more bandwidth to work on this.