The model is good at objects, especially detailed objects. It is a WIP, based on Wan2.2 t2v base model. It is, however, not very good at human (body/face) generation. Might need face-detailer for better faces. My very first model. Please let me know what you think.

Settings (for ComfyUI):

• 4 steps

• cfg 1

• res_2m/res_2s (sampler_name)

• bong_tangent (scheduler)

• some other settings might work, not tested.

Recommended prompt usage:

Add either of the following to your prompt to enhance subjects:

• portrait, 35mm lens, f/1.8, shallow depth of field

• street photo, 35mm lens, handheld, natural light

• cinematic scene, 50mm lens, f/2.0, bokeh

• ultra-wide 14mm lens, dramatic perspective, deep depth of field

Enjoy!

P.S.

• NSFW enabled, but not intended for such work!

• Tested on 12GB, 16GB and 24GB cards.

• Custom node recommended: RES4LYF (install from "ComfyUI Manager")

• For low VRAM (12GB) use ComfyUI-MultiGPU nodes (install from "ComfyUI Manager")