Do you get tired of the same body styles when generating NSFW in Leonardo? Tired of trying prompt variants? models?

Today I’m sharing a technique to get more toned, overflowing and sensual bodies.

Before we continue, let me quickly remind you that prompting should be the strong part. Just in case, here are some bits that may give you better results in models like Absolute Reality or DreamShaperv7.
Body mass: emaciated, skinny / slender, fat, exuberant
Body types: hourglass body type, triangle body type
Body details: slightly toned body, fit body, toned body, voluptuous

More Defined Bodies

Important: this technique uses Alchemy so the free tier of Leonardo.AI may not be enough.

In Leonardo.AI there are no LORAs. Yes, there are Elements which are implementations but nothing related to NSFW. But, there are other models, specifically Anime Pastel Dream which provides more exuberant and voluptuous bodies like the ones below.

But of course, these look like illustrations or hentai / manhwa. How can we make them look more realistic?

How to get images like the ones above? here are the sample prompts and the settings:

Settings (all use the same settings):

  • Model: Anime Pastel Dream
  • Alchemy = ON (not a free feature).
    • High Resolution = OFF
    • Expanded Domain = ON
    • Resonance: 13
    • Prompt Magic = OFF
  • Resolution: 728×728 (this is important)

Image 1 (Girl with long socks in bed)

Prompt: (((full body))) (((uncensored))) realistic, ray-traced, sensual uncensored and lewd gothic woman, triangle body type, short blue hair, touching herself lewd, pleasure face, pleasure expression, light blue eyes, wide pelvis, sweaty skin exalting her toned body, curvy body, striped long socks, triangle body type, ripped apart clothes, 8K ultra detailed illustration, high contrast, Manhwa style, background: bedroom, natural dawn light
Negative Prompt: chinstrap, face cover, face decoration, trash, blurred faces, blurry face, distorted face, weird, ugly, deformed, deformed faces, missing hands, deformed hands.

Seed: 678816768

Image 2 (Girl on the street)

Prompt: (((full body))) (((uncensored))) realistic, ray-traced, sensual uncensored and lewd Nier Automata’s 2B, triangel body type, wearing a full body white swimsuit, white hair, looking over her shoulder, big ass, back to the viewer, light blue eyes, wide pelvis, (thick hips), sweaty skin exalting her toned body, curvy body, 8K ultra detailed illustration, high contrast, Manhwa style, background: copied city, white city
Negative Prompt: chinstrap, face cover, face decoration

Seed: 17567232

Image 3 (Girl with overflowing body)

Prompt: (((full body))) (((uncensored))) realistic, ray-traced, sensual uncensored and lewd Nier Automata’s 2B, triangel body type, wearing a full body white swimsuit, white hair, light blue eyes, wide pelvis, (thick hips), sweaty skin exalting her toned body, curvy body, 8K ultra detailed illustration, high contrast, Manhwa style, background: copied city, white city
Negative Prompt: chinstrap, face cover, face decoration, trash

Seed: 431058432

If you are interested in how to create prompts like this and obtain NSFW results, check my other tutorial.

The Technique

To get more realistic results but keeping the body details from these pics we can combine these results with using another model and using an image as base.

This is the Image Guidance feature that in other AI tools is called Control Net. Basically reference images to help your prompt to get specific results. As usual, this is not exact, but it helps a lot! Worth mentioning also is that image 2 image (the one we will use) is not the only option, there are others like Normal maps, pose to image and more.

I’ll focus on Image 2 Image using the fist image of the girl with socks.

These are the results, for reference (and different strengths).

Steps to do this:

  • Download the original pic
  • Use the same settings (you can click on the three dots -> View Generation Info and click on Remix so Leonardo will copy everything)
  • Change these settings:
    • Model: DreamShaperv7
    • Resolution (on the left panel): as high as possible. This will prevent blurry faces and loose of details.
  • Go to the Image Guidance and use Image 2 Image and load the original pic to use as a guide.
    • Strength: this is the setting to play with: below 20 or above 60 may not work. In my experience 28 to 41 will do the job.

The higher the strenght the more the picture will look like the original.

You can also adjust the prompt if you want to change some things, but this will depend a lot on the strength so, as usual, it is better to work with a good original.

Final Notes

This is a solution to achieve more exuberant, toned, bulgy and marked bodies which I’ve found provides great control. Another way is to use the Elements option but these don’t work in all models.

Let me know what you think!