Need help with character consistency in ComfyUI
Hi everyone,
I’m trying to create consistent AI-generated photos of the same character using ComfyUI, but I’m struggling to find a reliable workflow.
My goal is to generate realistic photos of one character in different poses, outfits, camera angles, and environments while keeping the face and body consistent across images.
My setup
GPU: RTX 4050 Laptop GPU (6GB VRAM)
RAM: 32GB DDR5
OS: Windows
Software: ComfyUI Desktop
My main problems
I’ve managed to generate some good-looking portraits, but when I try to generate different poses or full-body images, the character’s appearance often changes.
Sometimes the face looks like a different person. Other times, the body proportions, facial features, or overall appearance become inconsistent.
I’ve tried different workflows and nodes, including face-related tools and multi-view approaches, but I haven’t found a reliable solution yet.
I don’t mind learning a new workflow or rebuilding my current one if necessary. I’d prefer to keep everything running locally rather than use paid cloud services.
What I’d like help with
What’s currently the best approach for maintaining character consistency in ComfyUI?
Should I stick with SDXL, train a character LoRA, use IP-Adapter or another identity-conditioning method, or try a different model?
How can I maintain both facial identity and body proportions across different poses and camera angles?
Is it realistic to achieve good results with only 6GB of VRAM?
Are there any recommended workflows or tutorials that actually work for this purpose?
I’m not necessarily looking for the most complicated or advanced setup. I just want a workflow that produces consistent, realistic results without having to fight the settings every time I generate a new image.
I’d appreciate any advice, workflow recommendations, or examples from people who have successfully solved this problem.
Thanks!