search

AUTOMATIC1111 / stable-diffusion-webui

Stable Diffusion web UI
GNU Affero General Public License v3.0
139.16k stars 26.41k forks source link

[Feature Request] Add GOBIG mode feature (allow upscale and adds details) #794

Closed mhnoni closed 1 year ago

mhnoni commented 1 year ago

txt2imghd is a port of the GOBIG mode from progrockdiffusion applied to Stable Diffusion, with Real-ESRGAN as the upscaler. It creates detailed, higher-resolution images by first generating an image from a prompt, upscaling it, and then running img2img on smaller pieces of the upscaled image, and blending the result back into the original image.

source code: https://github.com/jquesnelle/txt2imghd

Edit: ok after testing it, I'm not sure if it's worth adding it since it's very slow and the result wasn't great for me. since it splits the images into pieces, I'm not sure one prompts is going to be enough to get a good result since each peace needs different prompts like part of the image is a window and another part is a character, you can't apply one prompts for both sections.

devilismyfriend commented 1 year ago

SD Upscale on img2img is GoBig, it just needs to be added to the extras tabs.

pedropachecog commented 1 year ago

The new "High res fix" in txt2img is even better because it generates a low-res picture and runs SD again on it. You can even do it while still in latent space. It should alter the composition less than SD Upscale/GoBig.