Now this is exciting! Text based real image editing paper!

[bbecausereasonss]

Now THIS is exciting! > https://twitter.com/_akhaliq/status/1582175757153230849/photo/1

https://twitter.com/Buntworthy/status/1582307817884889088

https://github.com/justinpinkney/stable-diffusion/blob/main/notebooks/imagic.ipynb

[matrix4767]

This needs 30GB of vram, for 1 image to generate in 5 minutes.

[scorbon]

A100s and A6000s only for now. 4090s, too bad...

[bbecausereasonss]

Yes, but as usual this is the first hack implementation. Just wait a few weeks they'll optimize it and get it way down.

[hashnag]

You mean a few hours: https://github.com/ShivamShrirao/diffusers/blob/main/examples/imagic/README.md

[aoirusann]

Take a look: https://github.com/google/prompt-to-prompt
They say

The code was tested on a Tesla V100 16GB but should work on other cards with at least 12GB VRAM.

[angeleyeKJ]

Stable diffusion itself is said to need at least 10G vram... and it now runs on my old laptop without GPU. Just wait a few weeks.