Skip to main content
A prompt is a description of what you want the image to contain. Structure matters more than length — a well-ordered short prompt outperforms a long unstructured one.

Prompt structure

Write elements in this order, from most to least important:
PositionElementExample
1Subjecta woman in her 20s, red hair
2Action or poselying on a bed, looking at camera
3Settinghotel room, evening
4Lightingwarm side light, soft shadows
5Cameraclose-up, shallow depth of field
6Stylehigh detail
7Quality tagssharp focus, detailed skin
You don’t need every element. Start with subject and action — add the rest to correct specific problems with the output.

Good vs bad prompts

Bad: sexy woman, hot, beautiful, amazing photo, best quality This is all sentiment and no description. The model has nothing specific to work with. Good:

a woman in her late 20s, dark skin, black lingerie, sitting on a white bed, window light from the left, shallow depth of field

This gives the model subject, clothing, setting, lighting, and camera in one sentence.

What Prompt Enhancer does

PornX automatically enhances prompts. You can toggle it off in the settings panel if you want the output to follow your exact wording.

Negative prompts

A negative prompt tells the model what to exclude. Access it via the Advanced toggle. Common negatives: blurry, deformed hands, extra fingers, bad anatomy, watermark, text Don’t overload the negative prompt — five to eight terms is enough. More than that and the model starts ignoring all of them.

Tips

  • Put the subject first — the model weights earlier tokens more heavily
  • Be specific about body type, hair, clothing, and setting rather than using vague descriptors like “beautiful” or “perfect”
  • Add lighting terms to fix flat or overexposed results — soft natural light, dramatic side light, backlit
  • Use style and quality tags at the end, not the beginning
  • If you’re getting consistent artifacts, add the specific problem to your negative prompt

Advanced Prompting

Weights, negatives, and token order.

Model-specific tips

Prompting quirks per model.