20Loras
7/10/2025, 8:33:05 AM
No.11330128
>>11329933
It's just multiple different images put together, from the start. Like if the arm is bad on the image I like overall on image 3, but image 8 has a good arm, I copypaste it over. This workflow is best for the first image size because doing image2image with the same size blends things together nicely.
From there it's upscaling 2x and repeat the process if needed, just minor things like a hand or foot looking better.
Then another 2x upscale with low denoise and from there it's just going over the entire image with inpaint.
No controlnets, just bruteforcing it.
It's just multiple different images put together, from the start. Like if the arm is bad on the image I like overall on image 3, but image 8 has a good arm, I copypaste it over. This workflow is best for the first image size because doing image2image with the same size blends things together nicely.
From there it's upscaling 2x and repeat the process if needed, just minor things like a hand or foot looking better.
Then another 2x upscale with low denoise and from there it's just going over the entire image with inpaint.
No controlnets, just bruteforcing it.