Purpose

Is to generate a bit more realistic looking nude women as well as increasing the variety of output.

The main purpose is to show more natural looking breasts, genitals and nipples.

It does most of this, although it is weakest at genitals.

Prompting

This model was trained on descriptive prompts generated by JoyCaption Beta 1, with manual edits for accuracy and reducing wordiness. So descriptive prompts are best. See the images I uploaded for examples.

Trigger words

The model was not trained on a trigger word, instead using descriptive prompts to help guide the output. While there are no trigger words below are the most common tokens used in the training data.

Nude, topless and bottomless. These three token help guide the output to the appropriate lack of clothing.

Enhanced, round, perky, natural. Adding one or more of these words with breast will push the output towards the appropriate breast type.

Small, medium, large, very large. Adding one of these words will suggest the appropriate breast size. This is very finicky, but often describing body type helps.

Note: I also had tiny and flat chested, but for the most part Z-Image seemed to ignore this.

Skinny, thin, slender, fit. Describes the body type, and tends to aid in making the breast sized more accurate. Often you will have to add the word physique to this, buti it's not consistent.

Legs spread, spread wide. While I first attempted to use anatomical and descriptive terms for genitalia, but it seemed to make things worse. So I removed for the most part all references and just said legs spread and spread wide to show the vagina.