Stable Diffusion 3.5 Medium

I trained this model with a dataset of 10,000 images to improve the quality related to ???????, body proportions, and skin tone.

Normally, the SD3-3.5 model has issues rendering both hands properly — this model has been partially fixed (please use the skipLayer node as shown in my example).

This model does not require separate CLIP and VAE files. They are already integrated, so you can simply place the file in the "ComfyUI/models/checkpoints" folder and start using it.

Recommended settings:

Steps: 30–35

CFG: 4.5–5

Sampler: Euler + sgm_uniform

Note: This model requires 16–24GB of VRAM and 24GB of system RAM.