Whether the ipadapter tiled uses a square image or a rectangular image as a reference image, the resulting image will become fuzzy, not as clear as the image generated by ordinary ipaddapter apply, nor as clear as the image generated by direct use of model
You told me before that the use of tiled needs to match the composition, but after I tried this, I found that the generated picture is still a little fuzzy
In my screenshot, the three generation methods are, from top to bottom, only tiled, only composition, and tiled+composition
As you can see, the picture produced by composition alone is sharper
Whether the ipadapter tiled uses a square image or a rectangular image as a reference image, the resulting image will become fuzzy, not as clear as the image generated by ordinary ipaddapter apply, nor as clear as the image generated by direct use of model
You told me before that the use of tiled needs to match the composition, but after I tried this, I found that the generated picture is still a little fuzzy
In my screenshot, the three generation methods are, from top to bottom, only tiled, only composition, and tiled+composition
As you can see, the picture produced by composition alone is sharper