Spaces:

sd-concepts-library

/

stable-diffusion-conceptualizer

like 578

Running on A10G

I had the idea to try to teach the concept of images on white background, which then could be easily used for t-shirt designs etc.
So I trained with a few dozens of differnent png images. The model seems understand the concept of a white background, but any foreground objects it generates are pure "gibberish".

I am wondering if I should try again and just feed it a bunch of totally white images?