Thanks you

#1
by nicolollo - opened

Wanted to trying myself but didn't had time to , this model looks insane , you think is there a chance to successfully fine-tuning to output docci like captions ?

I think "more detailed captions" option is very close to DOCCI dataset level quality.

Yeah it surly is but lack of they docci keypoints that could enhance T2I like camera view for example, tho...it is still impressive for a model that small

gokaygokay changed discussion status to closed

Sign up or log in to comment