Add ability to use different loras without it merges with each other.

[CocoPuffLeo]

I really like the node you’ve created and have been experimenting with it for a while. So far, I’ve only been successful using a single LoRA containing multiple characters to generate images with different characters in it. However, when I try using multiple LoRAs with different characters, the results tend to fall apart (or should I say merge together) because they overlap before the sampling process.

I’ve tested several workflow approaches with this node by adding masks, using attention nodes with fingers crossed, tweaking settings and node configurations, but none have worked so far. It would be amazing if the node could apply each LoRA to its selected mask while keeping the overall image cohesive.

Right now, the only workaround I’ve found is to generate the base image first (when it looks kind of alright), then use inpainting in post-processing to apply details from each LoRA individually. It’s a very tedious process with a low success rate, but it’s the only method I know that somewhat works.

Thank you for taking the time to read this, and thanks again for creating this node. I look forward to hearing your thoughts soon!

[dcham23]

I've just installed this (Claude bot suggested it) and I was hopeful to find, at last, something to help with couples in (SDXL) image generation. However, as mentioned in the above post, the couples tend to merge into one freak of nature. And I can't get a hetero couple no matter what I prompt. Always two women. As @CocoPuffLeo says, the option of masking, inpainting and regenerating is tedious or worse - often a waste of time. This seems to be one area of AI image generation which needs a lot of work. I've tried with other models including Z-Image, Qwen and Chroma and can only manage a low percentage chance of getting a couple to interact naturally.