Mixing image generators, image prompts, and weird problems.
Google's Gemini does a pretty good job of imagining Mt Fuji as Vincent Van Gogh ...
... and Caspar David Friedrich
... but it flat out refuses to imagine them on display in a museum.
I tried both inputting the image, as well as rewriting the prompt.
Nope. Can't do it.
So I use Fooocus, a Stable Diffusion-based generator to get around these restrictions.
Straight-up using these images as prompts causes a re-imagining of the painting. Not what I want.
Easiest is just to outpaint and expand the borders.
Hmm. A bit too modern-art-gallery. After some fiddling, I could get exhibit label ...
... or a fancy period frame ...
... but somehow, not both.
Anyway, not a bad result after 1 hour fiddling, including generation time.
Comments
Post a Comment