Open yuesam opened 1 month ago
Nope, I think it's an error, in the huggingface version this lines is different so the prompt is automatically the same of the step1(https://huggingface.co/spaces/MohamedRashad/PaintsUndo/blob/main/app.py):
def interrogator_process(x):
image_description = wd14tagger.default_interrogator(x)
return image_description, image_description
hi - please do not change codes.
that huggingface space is wrong and gives worse results.
hi - please do not change codes.
that huggingface space is wrong and gives worse results.
Yes, but if that part is not correct... why step 3 must have a different prompt than step 1 prompt? (what prompt should be used in stage 3? default is 1girl, etc... so maybe a prompt more generic than prompt 1?)
Because the beginning frames are not correspondent to the tags - if your tags are "1girl, dog, cat", the beginning frames without that dog and cat will suffer. Later frames are less influenced, though.
@yuesam That's my mistake, i though the stage 1 prompt will enhance stage 3 but i was clearly wrong. I fixed it in the space, Happy building ^^
Do I change the prompt in step 3 to the prompt generated in step 1? Thanks