News for hackers. Stuff that matters.

A comment on Stable Video Diffusion. Parent comment

Re:Stable Video Diffusion by appplication on Tuesday November 21, 2023 @02:29PM (#38368705)
I don’t spend a lot of time keeping up with the space, but I could have sworn I’ve seen a demo that allowed you to iterate in the way you’re suggesting. Maybe someone else can link it.
[ Parent | on HN ]
Re:Stable Video Diffusion by ssalka on Tuesday November 21, 2023 @02:48PM (#38369040)
My guess is you're thinking of InstructPix2Pix[1], with prompts like "make the sky green" or "replace the fruits with cake"

[1] https://github.com/timothybrooks/instruct-pix2pix

[ Parent | on HN ]
Re:Stable Video Diffusion by appplication on Tuesday November 21, 2023 @05:33PM (#38371395)
This is exactly it!
[ Parent | on HN ]
Re:Stable Video Diffusion by tjoff on Tuesday November 21, 2023 @03:22PM (#38369515)
Emu-Edit is the closest I've seen.

https://emu-edit.metademolab.com/

https://ai.meta.com/blog/emu-text-to-video-generation-image-...

[ Parent | on HN ]
Re:Stable Video Diffusion by accrual on Tuesday November 21, 2023 @02:45PM (#38368989)
It's not exactly like GP described (e.g. move bike to the left) but there is a more advanced SD technique called inpainting [0] that allows you to manually recompose parts of the image, e.g. to fix bad eyes and hands.

[0] https://stable-diffusion-art.com/inpainting_basics/

[ Parent | on HN ]