This LoRA will generate photorealistic women with a certain consistency to their stomach shape and navel. Note that I am making several more LoRAs grouped by navel shapes and belly details. I found that even with good prompting, results are much better if you keep the training set focused on these particular details in isolation to avoid bad crossfading between concepts.
I found a small amount of women whose navels had stretched out to the point where they basically became flat slightly dark patches of skin. Don't worry, there will be further LoRAs with different appearances coming soon.
Basic Usage Notes
I recommend using it with a strength of 0.75, with res_2s sampler and ddim_uniform scheduler.
The prompt isn't very strict and not based on language that is atypical for English prompts - but you can refer to the example images for some inspiration.
Things like a pregnant woman, has stretch marks on her growing belly... etc etc. There is quite a range of descriptors.
Description
FAQ
Comments (5)
Thank you very much for your time and effort making this, it is absolutely fucking great, I'll post some images when I get chance
which workflow did you use?
I used a very basic ComfyUI node setup to load the model, load the lora at 0.75 strength, with res_2s and ddim_uniform. There was nothing additional being used to generate these images
Just want to add btw, I added this data to my Z-Image Spice checkpoint, so it is finetuned on the dataset from this LoRA too
Have you used this lore with the regular base version of Z image?