Closed enn-nafnlaus closed 1 year ago
thx for the suggestion, I guess that you are working with the img2img pipeline? When img2img, use that "Inpaint upload" instead of ”img2img" will literally do the travelng with one given mask. For txt2img, I could not find an understandable reason to use mask, since the generated content is almost unpredictable.
btw, I'm not quite familiar with that "Custom depth images" script, do you mean this or something else? Any links please? btw again, ptravel mainly works for txt2img, performance is generally worse in img2img (it more like a seed_travel under img2img
Thanks for the response :) No, I'm working with txt2img.
Yes, that's the extension. It should make it clear how they pull it off. Unfortunately I can't run that script and yours at the same time, as you can only run one script at once!
Custom depth maps let you pull off stuff like this:
Being able to prompt travel, but with a custom depth map, should make some really neat effects!
🆗 I've got the idea , but it seems to be some new functionality from SD2...?
Where could I get a depth2img model ckpt for the development & test?
So that I could add the very basic feature that depth-image-io-for-SDWebui
does.
The depth2img checkpoint is here:
https://huggingface.co/stabilityai/stable-diffusion-2-depth/tree/main
Nice, I'll make it
Awesome! :)
Awesome!!!
See the "Custom depth images" script for an example. It's a shame that one can't both provide a custom depth map, and run prompt travel, at the same time.