It reminds me of a kind of shower where the woman changes for hot or cold.
That's amazing! It also added a bathing suit to keep it SFW
I recently found You are not allowed to view links.
Register or
Login to view., which mixes two images taking style from the first and content from the second. It clearly works better when input and output are photographs, but in many ways it is quite impressive. All parameters were left at default values, with the exception of "guidance scale 12" which (if I understand correctly) makes the output closer to the content image.
Style images used:
[
attachment=9367]
Prompts:
scholar in his study, Renaissance painting
scholar in his study, sci-fi movie
scholar in his study, post-modernist drawing
woman with alien machinery, Renaissance painting
woman with alien machinery, sci-fi movie
woman with alien machinery, post-modernist drawing
crossbowman with white hat, Renaissance painting
crossbowman with white hat, sci-fi movie
crossbowman with white hat, post-modernist drawing
[
attachment=9366]
[
attachment=9365]
[
attachment=9368]
That's a really cool tool. It seems to leverage the capabilities of this kind of image generators in the right way.
(15-05-2022, 01:36 PM)MarcoP Wrote: You are not allowed to view links. Register or Login to view.These "manuscript pages" were created by You are not allowed to view links. Register or Login to view. on the basis of the prompts:
["A page from the Voynich manuscript", "ancient manuscript"]
The following animated gif shows some of the processing steps. If I understand correctly, Disco Diffusion starts from pure noise; each step removes some noise, generating a better fit for the verbal prompt. The creation of a language/image mapping (the method used here is called CLIP, described in You are not allowed to view links. Register or Login to view.) is possible thanks to the millions of images with a corresponding text that are available on the internet.
This is hilarious! Maybe AI isn't so bad after all, at least in terms of making things for people to laugh at without that being the intent. Hoo boy..
What's interesting about this thread is that it allowed relative laymen like me to kind of follow along with more rudimentary forms of image generation.