#concatenation is selecting either the style prompt or the base prompt, not combining as expected
12 messages · Page 1 of 1 (latest)
When you say it 'only uses the style prompt' - Are you saying that that's what is showing up in the metadata, or that your subject isnt showing up
so if I concatenate, and my positive prompt says "tortilla chips" and my style prompt says "melty cheese" I get images of cheese. If I do not concatenate, I get images of tortilla chips, and no cheese. In no case does it make nachos.
If you don’t concatenate, you need to include tortilla chips in the style prompt as well.
It’s a bit of a misnomer.
if I don't concatenate, the style prompt box doesn't even show up - hence why am confuzed
Right. You have two options - concat, or don’t. The hybrid option from before is no longer available
if I use concatenate and have to put everything in the style prompt, or I don't and put everything in the positive prompt, what's the point of retaining the concatenate function?
I'm not exactly sure what you mean
Concatenate hides the style box and copies everything over - Consider this "Simple prompting mode". Turning it off exposes both text encoders letting you do as you wish (same as the old UI when concat was off)
The fact that you have to pass in the subject into both encoders is a reality of how the model was trained
hm. I understand what you're saying - it's just that I wasn't doing that in the release version of Invoke and it was working the way I describe it; with concat on both prompts were affecting the output, I wasn't including the subject in the style prompt. at least that's what my outputs were leading me to believe. That wasn't actually true, it means I write better prompts than I thought 😄 all good, can close this out - I understand what's happening now.