Commit History
-
Correct `sample_size` of Stable Diffusion 1's unet to have correct width and height default (#3)
- Correct `sample_size` of Stable Diffusion 1's unet to have correct width and height default (25f3c34703421b79db4b4d19ac1bd0d0082b7af7) Co-authored-by: Patrick von Platen <patrickvonplaten@users.noreply.huggingface.co>f54964b
-
Add Diffusers weights (#2)
- Add Diffusers weights (555ee92053e80fdf6ad60267ed08b4972141ac5d)9ea3be5
-
Update README.md
54b340a -
Upload mjstyle.ckpt
ca45f25
-
Update README.md
daa9823 -
Update README.md
I have seen a bunch of MJ Models flying around, but I couldn't get the --test aesthetic from them. So I went ahead and collected 35 images from midjourney gallery, flipped them, and trained a model for 7000 steps (640x640). Results were surprisingly pretty good and pretty customisable. token used - mjstyle Adding it to the end of the prompt gives best results. E.g.:- painting of a robotic snake, 4k, hd, by artgerm, by greg rutkowski in mjstyle, (Modular Origami) [ Euler a, CFG scale: 7, Seed: 4077618648, Size: 640x640 ]  painting of emma watson, 4k, hd, by artgerm, by greg rutkowski in mjstyle [ Euler a, CFG scale: 7, Seed: 4275282758, Size: 640x640 ]  painting of a cute panda in woollen sweater, pastel colours, unreal engine, volumetric lighting, oc…eee78cc -
initial commit
ff9c07d