This LoRA is intended to increase the variety of toe/foot poses (e.g. scrunching toes, spreading toes, etc.).
The LoRA should respond to natural language prompts.
🚨 Important 🚨: Recommended LoRA strength is 0.3 (I may have messed up the conversion to the ComfyUI .safetensors format and the LoRA is way too strong at a strength of 1.0)
Training
The training images were rendered by Grok Imagine (Quality) based on pose references extracted from real photos (using my Mannequins LoRA) against a clear white background with minimal clothing. This creates a consistent aesthetic and simplifies captioning considerably.
Description
I trained this one on fal.ai and had to convert it to ComfyUI manually. The scaling may be a little wack, but it seems to work as intended at a strength of around 0.3.
Training data was generated by Grok Imagine, based on foot pose references made with my "Mannequins" LoRA for Qwen-Image-Edit.
FAQ
Comments (3)
Krea 2 is so fucking crazy 🤤
Watched ComfyUI's presentation on it earlier this week... and I certainly didn't expect to be like this
Choosing the right model used to be like choosing between apples and oranges. Now it's like choosing between cologne or perfume, and then which brand you want, and how much you want to spend.
fractal backrounds its a hack?
Good lora!