← sources

5 character LoRAs in one image?

source: reddit (reddit:StableDiffusion)published 2026-11-18 20:58ingested 2026-11-19 02:37

retained text

I've been trying to get five distinct characters into one Anima image, each with their own character LoRA. I thought I'd share my results:

1. Prompts only, all five LoRAs at 0.6. Features bled everywhere: the cat girl got another character's elf ears and braids in all 4 seeds, and the dark-skinned character never showed up. 2 of 5 right.

2. Sen-sou's Anima regional conditioning (attention masks, LoRAs still global). Only 3 women drawn, every seed. Still 2 of 5.

3. RS MultiCharacter Sampler with a LoRA per region, using my original overlapping boxes. The face where three boxes overlapped came out as a blend, and I got extra women in the background.

4. Same sampler, boxes that don't overlap, plus its per-character face pass. 4 women every time and 3 of 5 right in 3 of 4 seeds. First setup where the dark-skinned character appeared at all. But two boxes stacked in one corner always merged into a single woman.

A separate face pass nudged details like eye color and lips, but couldn't bring back someone who never got drawn. Strong character LoRAs also override the outfits you prompt. And it's slow: 7 to 8 minutes an image, since every LoRA region gets its own prediction each step.

Has anyone gotten overlapping characters (someone sitting in front of someone else) to keep their own LoRAs on Anima? Or should I figure out how to generate them separately and composite?

document fetched 2026-11-19 00:31 · canonical https://reddit.com/r/StableDiffusion/comments/1xzqnav/5_character_loras_in_one_image