[V2.0 Update ]
๐บ๐ธ English
This workflow uses the ImageCompositeMasked node instead of a Stitch node, ensuring that only the inpainted area is affected while the rest of the image remains untouched. The structure has been simplified for ease of use.
The Save Image node at the bottom right produces the final result.
๐ฐ๐ท ํ๊ตญ์ด
์ด ์ํฌํ๋ก์ฐ๋ Stitch ๋ ธ๋ ๋์ ImageCompositeMasked ๋ ธ๋๋ฅผ ํ์ฉํ์ฌ,
์ธํ์ธํธ ์์ญ๋ง ์์ ๋๊ณ ๋๋จธ์ง ๋ถ๋ถ์๋ ์ํฅ์ด ๊ฐ์ง ์๋๋ก ๊ตฌ์ฑํ์ต๋๋ค.
๋ํ ์ฌ์ฉ์ด ๊ฐํธํ๋๋ก ๊ฐ์ํํ์์ต๋๋ค.
์ฐ์ธก ํ๋จ์ Save Image ๋
ธ๋๊ฐ ์ต์ข
๊ฒฐ๊ณผ๋ฌผ์ ์ถ๋ ฅํฉ๋๋ค.
[contact]
Youtube : https://www.youtube.com/@A01demort
Naver Cafe : https://cafe.naver.com/sdfkorea
#์คํ
์ด๋ธ AI ์ฝ๋ฆฌ์ #A1 #AI๋ ์์ด์ #๋ํ๋ฏผ๊ตญ AI ์ปค๋ฎค๋ํฐ
#A01demort
Description
FAQ
Comments (4)
Is there any reason not to use the official Flux tools inpaint model?
Nah, no special reason.
Just turned one of the many methods into a workflow.
I know you tried to make it simple, but you did not composite in the end after vae decode, so this will degrade the whole image with every inpaint. https://new.reddit.com/r/StableDiffusion/comments/1gy87u4/this_looks_like_an_epidemic_of_bad_workflows/
Thx for enlightening me! Iโm going to fix this workflow soon.

