enhancement ideas

[MoreColors123]

This works wonderfully. Thank you very much for your work! The regional prompt extension in A1111 has alyways been a mess, UI-wise. Very good idead to add Lora-functionality, so one can use LCM-sampler too, which generates pretty very fast.

Anyway, I can think of the following suggestions/enhancements:

1. I would absolutely love to be able to just right-click and copy the generated image to the clipboard. Or via some button. Is it possible to add that feature easily?
2. Be able to edit the size of boxes after setting them. A transformation tool like in photoshop or something, which can move the box too, would be great.
3. Set the full canvas as a background-box with a button; so i can e.g. set the background as "jungle" and add objects like "monkey" onto it. (If that is possible?)
4. Edit the width and height of the whole canvas (768 x 512 and vice versa would be great)

And one question: what is the Canvas Backdrop supposed to do? i loaded a picture and generated something on it in a box, and the loaded picture didn't appear in the result in any way.

[Acephalia]

Added fixed 512x768 support in this fork.

Edit: As advised by @mut-ex expect bleed and cloning. I’ve done a bit of testing on my end as well and it is perfectly usable for for simple compositions. Complex/overlaps not so much.

[mut-ex]

1. Sure that won't be a problem to implement
2. Since I developed the whole "drawing" framework from ground up; it will take some amount of work to add box moving and resizing capability. I'd still like to do that at some point.
3. That's a nice suggestion! But also have you tried to put "jungle" in the positive base prompt? I wonder if it will have the same effect. I will also look into this.
4. I can add that but in my experience GLIGEN gets pretty unpredictable when you deviate from 512x512.

Since ComfyUI doesn't currently have image2image implementation of GLIGEN, the set canvas backdrop is currently just so you can load a reference image to help you while you draw the boxes.