I feel there is not enough good anima loras for realistic, my rig is too slow to generate with Krea2 so I'm proud to present my (not perfect) instagram baddie lora. Results are... better than I expected, trained on Civitai.



Trained on 133 hand-picked photos mix of framings: 24 close-ups, 32 upper body, 30 cowboy shots, 35 full body and 12 from behind.

Only after the upload I've noticed the prompts are not visible.... cba fixing that so if u want prompts + whole workflow download the image and put it in comfyui. Though FYI my workflow has a face swap mode too which is disabled on default but is very neat

P.S. If you have any tips to improve this lora feel free to DM me.