Do not connect VAE to the text encoder for perfect 1:1 edits!
#26
by
CamiloMM
- opened
There's a bit more to it, especially when it comes to having multiple input images, but the gist of it is that you can get perfect 1:1 edits with no pixel shift, unzooming, etc! It's just that the ComfyUI devs did goofed and made the official workflow wrong.
Take a look at this post where I learned of it:
https://www.reddit.com/r/StableDiffusion/comments/1o01e6i/totally_fixed_the_qwenimageedit2509_unzooming/
Might need a few tweaks to the workflow, but totally worth it (they do share a workflow, a bit convoluted for my taste, but I took the bit that mattered into my own).
CamiloMM
changed discussion title from
Do not connect VAE to the text encoder!
to Do not connect VAE to the text encoder for perfect 1:1 edits!
I ended up changing the text encoding node to remove scaling (unless the image is really big). There are some good workarounds though.