I highly recommend playing with stable diffusion.
Images presented below were generated using image to image translations guided by text prompt. Source images were captured by me on various occassions. Img2Img requires some patience to get results that can satisfy your taste, my recommendation for initial setup (other stuff leave at defaults):
{
strength = 50 /* balance between text input and your images */
ddim_steps = 60 /* no strong benefits with increased value, for sure use ddim sampling */
scale = 7.5 /
}
Dimensions ~ 1344x960 are possible with at least 12GB VRAM in case of GPU/cuda processing.
Super easy tutorial how to setup SD can be found here https://rentry.org/voldy
"prompts": abstract art painting, picasso geometric painting. kandinsky style painting
