Does anyone have any tips for using multiple characters on an image without having them blending to together.

I use civitai and stable diffusion.

gtgtgtg said:
Does anyone have any tips for using multiple characters on an image without having them blending to together.

I use civitai and stable diffusion.

unfortunately most of the techniques people usually use to avoid that problem (like regional prompting) probably aren't an option when you're generating on CivitAI. SDXL models are probably the best bet if you're limited to only prompting, it really depends on how complex a scene you're trying to create and the biases the model has.

gtgtgtg said:
Does anyone have any tips for using multiple characters on an image without having them blending to together.

I use civitai and stable diffusion.

You'll have to bruteforce through models and see which are bleeding less. If you're adamant on not using other tools, that's probably your only option.

gtgtgtg said:
Does anyone have any tips for using multiple characters on an image without having them blending to together.

I use civitai and stable diffusion.

noobai and its variants are actually pretty good at separating named characters, i've also done generic anthros and had them separate but that is a lot less consistent than using a named subject. otherwise like others have said you'd have to use methods available to local generation.

aislopper said:
noobai and its variants are actually pretty good at separating named characters, i've also done generic anthros and had them separate but that is a lot less consistent than using a named subject. otherwise like others have said you'd have to use methods available to local generation.

So basically only use stable and ill have to use line breaks
Well that sucks

gtgtgtg said:
So basically only use stable and ill have to use line breaks
Well that sucks

Ideally you want to gen locally, but if the models I mentioned aren't usable on civit you can try browsing tensor.art as a backup online gen site.