search

jamriska / ebsynth

Fast Example-based Image Synthesis and Style Transfer
https://ebsynth.com
1.49k stars 195 forks source link

fyi #32

Open johndpope opened 1 year ago

johndpope commented 1 year ago

https://twitter.com/karenxcheng/status/1565003860854185984?s=20&t=4_P3UToO3QdyMj1BxuRRQQ

shodanx2 commented 1 year ago

Funny you should mention that, I was thinking, masking each frame of a video is a huge pain in the rear, how about using that thing that Joel Haver uses to make his video. Basically you draw one still and it morphs that style with video motion until the morph gets too degraded.

So, next question , how do we get this ebsynth software with a web ui in a container and automated the process as much as possible back and forth out of StableDiffusion

I guess it's time to ask the people over at https://github.com/sd-webui/stable-diffusion-webui

johndpope commented 1 year ago

next level - https://twitter.com/LighthiserScott/status/1571351893208715264?s=20&t=TvJBRPRAGQKPfnx1F67ARA

shodanx2 commented 1 year ago

It's pretty great, I was thinking of erasing a part of the image and then use the input video to animate the erased part So, from the first video, when he erases the inner part Then the transparent region is morphed, maybe with ebsynth over all the other frames, then the AI fills in the blank region frame by frame in img2img mode