This workflow aims at creating images 2+ characters with separate prompts for each thanks to the latent couple method, while solving the issues stemming from it. Those include inconsistent perspective, jarring blending between areas and inability to generate characters interacting with each other in any way. All of those issues are solved using the OpenPose controlnet, though of course it's not as simple under the hood. You can, though don't need to provide your own OpenPose images, you can extract them from photos or generate them on the spot. The main limitation is how well the preprocessor can detect the pose in the image. For that reason providing your own image will still give the best quality, but of course requires more effort.

The workflow includes a second pass, an upscaler and many QoL features.

It is set up to work with both SD1 and SDXL support, but SDXL only works properly for pregenerating images for pose extraction. The reason is not for the workflow itself, but rather subpar performance of the so-far existing OpenPose ControlNets for SDXL, at least among the ones I found. If something that works better is released, this workflow should be ready for it.

You can technically use character LoRA with this. It does not limit their influence by area, but results are decent as long as they require trigger words. All-in-One type LoRA have the most potential.

More specific instructions inside.

Requires:

• WAS's Comprehensive Node Suite - ComfyUI
• ComfyUI Impact Pack
• ComfyUI Inspire Pack
• ComfyUI-Custom-Scripts
• comfyui_controlnet_aux

(with ComfyUI Manager go to Manager -> Install Missing Nodes)