Generates images of the Sultan from Aladdin, SFW or NSFW.

Drop any sample PNG onto the ComfyUI canvas to load the workflow used to generate the image.

0.80 LoRA strength allows me to pull NSFW out; otherwise, it becomes more difficult.

Positive prompt structure:
- Danbooru tags in comma-separated sequence (subject (place explicit 5u1t4n features here, e.g. white beard, white mustache, elderly man), rating (e.g. nsfw, explicit), body/anatomy, pose, camera/viewpoint, setting, lighting). The first tag should be the training tag 5u1t4n, but I did not always do this. A period and carriage return at the end of this group separates it from the following natural language prompt.
- Natural language prompt describing the image (I have been using 1-2 sentences).

Negative prompt:
Score tags: score_1, score_2, score_3; meta tags, e.g. artist name; quality tags: blurry, jpeg artifacts, chromatic aberration. I have also been adding tags into the negative prompt like 'dark beard' or 'clean shaven' to elicit Sultan's white beard. I also had some trouble with wayward/rogue eyes, so I would add 'empy eyes' to make them come through.

CLIP: qwen_3_06b_base.safetensors
VAE: qwen_image_vae.safetensors