VQGAN paper:
https://compvis.github.io/taming-transformers/
Colab:
https://colab.research.google.com/github/eps696/aphantasia/blob/master/CLIP_VQGAN.ipynb
Example:
"a dark fantasy world rendered in unreal engine"
"colorful nebula forest"
"Frosted mountains, captured with sony alpha camera, unreal engine, rtx"
View attachment 288220
"GBAtemp flag" these ai don't just generate one image, they generate multiple and refine the image over time, here is the progression:
View attachment 288223
Rudalle:
https://rudalle.ru/en/
Colab:
https://colab.research.google.com/drive/1wGE-046et27oHvNlBNPH07qrEQNE04PQ?usp=sharing (you will have to translate any text to russian first since it was trained in russian lol)
Examples:
"Flat Minimalist House"
View attachment 288218
View attachment 288219
The model below doesn't work the same way the others do, instead you train it with a dataset of a lot of similar images and it tries to generate similar ones... it's the model used in
https://thispersondoesnotexist.com/
Stylegan2:
https://developer.nvidia.com/blog/synthesizing-high-resolution-images-with-stylegan2/
Demos:
https://thisxdoesnotexist.com/ and
https://github.com/justinpinkney/awesome-pretrained-stylegan2
My personal favourite model:
https://github.com/Vilagamer999/SneakGAN (totally not by me I swear)
Examples:
Generated on a dataset of cat images:
View attachment 288221
Generated on a dataset of self portraits:
View attachment 288224
Generated on a dataset of cars:
View attachment 288225
And finally the best one (dedicated to
@WiiMiiSwitch):
View attachment 288226
Enjoy
