Try single asset training

MountaintopLotus / braintrust

A Dockerized platform for running Stable Diffusion, on AWS (for now)

Apache License 2.0

1 stars 2 forks source link

Try single asset training #13

Open JohnTigue opened 1 year ago

JohnTigue commented 1 year ago

Nerdy Rodent, one of the best SD YouTubers, has a video, Textual Inversion and Hypernetworks - Stable Diffusion 2, wherein he gets some pretty good (not perfect) results by training SD on a single image. Our art is just a set of individual unrelated images. Let's see what happens. (Requires moving to AWS to train on a decent GPU).

JohnTigue commented 1 year ago

Wow, 3D animations with 4 training images: My friend makes this video using 4 hand-drawn images + #Live3D-v2 + #stablediffusion.

JohnTigue commented 1 year ago

Saw someone talking about needing only 4 or 5 images.