Does the thing it says. 0.4-1. Smash the like button if you want to see larger version with variants trained as separate concepts.
Description
A test run that works pretty well. Text to image and i2i both work. Prompt for crudely drawn MsPaint illustrations, insane characters, silly faces etc. Doesn't have good character knowledge.
FAQ
Comments (5)
Billions must inference
Did you train it on pairs or single images?
Single images. I'm not sure if there's any advantage for pair training for this model
coal example images
I admit, had to test it doesn't ruin text generation.