ComfyUI and QWEN have a few issues regarding image creation. Stitching two images together alone does not fully work, and there has been some confusion regarding it.
This is likely a fix for a temporary problem - but I still wanted to offer my 'optimized' workflow in an attempt to help everyone get it earlier.
Description
FAQ
Comments (5)
It is known that Qwen is sensitive to the output rez when compared to input rez with regard to complex operations like stitching. So the solution is to ensure an aspect ratio that matches the prompt requirement.
That's been my experience as well! Essentially all I did was re-size it. If Image 1 and Image 2 are grossly different - especially if one is tall, and the other wide - it still tends to break things.
Right now my only goal is to find a generic way to handle 'most cases'.
There a new (today) Qwen-Image-Lightning 8 step lora specifically for Qwen-Image-Edit you might want to try if you haven't already.
Dang. That's hot off the press. Thank you!! ill give it a shot!!
Please mention stuff like "Require sage attention"