Body type prompts describe the build of an adult (18 and over) fictional AI character: slim, athletic, curvy, plus size, and more. Use clear adult anchors such as mature woman or grown adult, tasteful tag-style descriptors, and a baseline safety negative on every prompt. Never use youthful, teen, or any age-regression phrasing. All subjects are fictional and AI-generated.
Getting a consistent, intentional body type out of a diffusion model is one of the trickier parts of NSFW prompting. The model averages across enormous amounts of training data, so without clear direction it drifts toward a generic default. The fix is a small, deliberate vocabulary of adult body descriptors combined with explicit adult anchoring. This guide gives you that vocabulary, a copy-paste library per build, and a strict safety framework so every character you generate is unmistakably an adult.
Before anything else, the non-negotiable rule for this entire topic: every character is an adult, 18 or older, fictional, and AI-generated. We never use words like young, youthful, teen, schoolgirl, petite-as-age, loli, or any other phrasing that suggests a minor or a minor-appearing subject. Body descriptors are about adult build only. Every prompt in this guide carries the baseline safety negative child, minor, underage, loli, shota so the model is steered firmly away from any underage output. If a render ever looks ambiguous in age, discard it and re-prompt with stronger adult anchors.
Why body type prompts drift
There are two reasons body type comes out wrong. First, no anchor: if you only write a single descriptor, the model has nothing to lock onto and reverts to its default. Second, conflicting signals: descriptors that pull in opposite directions, like asking for both slim and voluptuous without a clear hierarchy, leave the model to guess.
The solution is the same structure every time. Lead with a clear adult subject anchor, then add one primary build descriptor, then at most two supporting feature descriptors. That gives the model a strong, unambiguous target. If you want to understand the broader prompt structure this sits inside, our NSFW AI prompt formula walks through subject, build, pose, setting, and lighting in order.
Adult anchoring: the part you never skip
An adult anchor is a phrase that tells the model the subject is a grown adult. It does two jobs: it pushes the output toward adult proportions and faces, and it reinforces the safety negative. Good anchors include adult woman, adult man, mature woman, grown adult, woman in her late twenties, 30 year old man, and an explicit numeric age that is clearly adult such as 25 years old.
Use at least one anchor in the positive prompt of every generation. On models that respond to it, an explicit adult age plus a maturity word (mature, grown) is the strongest combination. This is not optional framing, it is core craft for responsible adult work, and it produces better, more consistent adult characters as a side benefit.
The adult body type vocabulary
Here are the descriptors that reliably steer build, grouped by category. All describe adult physiques only.
Overall build
slim, slender, lean, athletic, toned, fit, muscular, curvy, voluptuous, hourglass figure, plus size, full figured, soft build, average build. Pick one as your primary build descriptor.
Supporting features
broad shoulders, narrow waist, wide hips, long legs, defined abs, strong arms, soft midsection. Add at most two so the model is not overwhelmed.
Skin and tone
natural skin texture, realistic skin, tan skin, fair skin, freckles, glowing skin. These help realism and read as clearly adult, lived-in skin.
Height and frame
tall, petite frame (frame only, never age), statuesque, compact athletic build. Note that petite refers to frame size for an adult, and must always sit beside an adult anchor and adult age so there is zero age ambiguity.
Body type comparison table
| Build | Primary descriptor | Good supporting features | Pairs well with checkpoint |
|---|---|---|---|
| Slim | slim, slender | long legs, narrow waist | SDXL realistic, Flux |
| Athletic | athletic, toned | defined abs, strong arms | Pony, SDXL |
| Curvy | curvy, hourglass figure | wide hips, narrow waist | Pony, Illustrious |
| Voluptuous | voluptuous, full figured | soft build, glowing skin | Pony, realistic SDXL |
| Plus size | plus size, full figured | soft midsection, wide hips | SDXL realistic |
| Muscular | muscular, strong | broad shoulders, defined abs | SDXL, Pony |
Copy-paste body type library
Every example below is an adult (18 and over), fictional, AI-generated character. Each leads with an adult anchor, carries one primary build descriptor, and includes the baseline safety negative. Swap the build descriptor to retarget. Keep the adult anchor and the safety negative on every single render.
Slim adult character (SDXL realistic)
photo of an adult woman, 27 years old, mature, fictional AI character,
slim slender build, long legs, narrow waist, natural skin texture,
soft studio lighting, editorial photography, sharp focus
Negative: child, minor, underage, loli, shota, youthful, teen, lowres,
bad anatomy, deformed, extra limbs, watermark, textAthletic adult character (Pony Diffusion)
score_9, score_8_up, score_7_up, 1woman, adult woman, 28 years old, mature,
fictional character, athletic toned build, defined abs, strong arms,
confident pose, gym setting, photorealistic
Negative: child, minor, underage, loli, shota, youthful, teen, bad anatomy,
lowres, watermark, deformed hands, extra limbsFor more on Pony scoring tags and how build descriptors interact with them, see our Pony Diffusion guide.
Curvy adult character (Illustrious / anime)
masterpiece, best quality, 1girl, adult woman, mature, 26 years old,
fictional character, curvy hourglass figure, wide hips, narrow waist,
detailed eyes, soft lighting
Negative: child, minor, underage, loli, shota, youthful, teen, worst quality,
lowres, bad hands, extra digits, watermark, signatureThe Danbooru tags reference and our Illustrious model guide cover the anime-specific descriptors in more depth.
Plus size adult character (realistic SDXL)
photo of an adult woman, 32 years old, mature, fictional AI character,
plus size full figured build, soft midsection, wide hips, glowing skin,
natural skin texture, soft window light, shallow depth of field
Negative: child, minor, underage, loli, shota, youthful, teen, plastic skin,
lowres, deformed, extra limbs, watermark, textMuscular adult male character (Flux)
Portrait of an adult fictional man, early thirties, muscular athletic build
with broad shoulders and defined abs, natural skin texture, soft studio
lighting, shot on film, tasteful and refined.
Negative cue: child, minor, underage, loli, shota, youthful, teen, clearly an
adult, no deformed anatomy, no watermark.Flux reads natural language rather than tag soup, so describe the build in full sentences. Our Flux NSFW guide explains its prompting style and how it handles negatives.
Want to test a build right now? Paste any recipe into our free generator and change only the build descriptor to see how each one reshapes the same adult character.
Controlling build with weighting
If a build descriptor is being ignored, you can emphasize it with attention syntax on AUTOMATIC1111, Forge, and ComfyUI. For example, (athletic build:1.2) raises the weight of that phrase. Keep weights modest, between 1.1 and 1.3, because pushing too hard distorts anatomy. Our prompt weighting guide covers the exact syntax and the differences across interfaces, including why Flux handles emphasis differently.
Note that weighting only affects how strongly a phrase is applied. It does not change the meaning, so it can never be used to override the adult anchoring or the safety negative. Those stay fixed regardless of any weights you add.
Keeping a body type consistent across images
Build often shifts between generations even with the same prompt, because the seed changes. To lock a character’s body type across a set, you have a few options. Reusing the same seed keeps proportions stable. Reference based tools hold the figure steady: our IPAdapter guide shows how to carry a character’s look across images, and the broader character consistency techniques article compares every method including LoRAs and embeddings.
For pose-locked consistency where you want the same build in a specific position, OpenPose pose control and the general ControlNet guide let you fix the skeleton while you vary other attributes. Combining a stable seed, a strong build descriptor, and a reference tool gives you the most reliable adult character continuity.
Multiple characters with different builds
When a scene has two adult characters with different body types, a plain prompt tends to blend their features. Regional prompting solves this by assigning each descriptor to a separate area of the canvas. Our Regional Prompter guide explains how to keep, for example, one slim adult character and one curvy adult character distinct rather than averaged together. Apply the adult anchor and safety negative to the whole prompt, and the build descriptors to each region.
Common body type mistakes
The descriptor is ignored. You probably have no adult anchor, so the model defaults. Add an explicit adult age plus a maturity word, then your build descriptor.
Conflicting builds. Asking for both slim and voluptuous without hierarchy confuses the model. Choose one primary build and let supporting features refine it.
Distorted anatomy from over-weighting. Weights above roughly 1.3 warp proportions. Pull the weight back and add anatomy negatives like bad anatomy, deformed, extra limbs.
Age ambiguity. If a render looks anything other than clearly adult, do not use it. Strengthen the adult anchor, add youthful, teen to the negative alongside the baseline tokens, raise the stated age, and regenerate. This is the most important fix on the list.
| Problem | Likely cause | Fix |
|---|---|---|
| Build ignored | No adult anchor | Add adult age plus mature |
| Mixed-up build | Conflicting descriptors | One primary build only |
| Warped anatomy | Over-weighting | Lower weight, add anatomy negatives |
| Age looks unclear | Weak anchoring | Strengthen anchor, discard and re-prompt |
Using LoRAs to lock a build
When a build descriptor in the prompt is not enough, a LoRA can lock a specific physique far more reliably than words. Body-focused LoRAs bias the model toward a particular adult build, and they stack with your prompt descriptors. The advantage is consistency: the same LoRA at the same weight produces the same adult build across many images, which is hard to achieve with text alone. Our best NSFW LoRAs roundup covers what is available and how to choose one.
Keep LoRA weights moderate, usually between 0.5 and 0.9, so the LoRA shapes the build without overpowering the rest of the prompt or distorting anatomy. A LoRA at full strength can flatten variety and introduce artifacts. As always, a LoRA changes appearance, never the safety rules: the adult anchor and the child, minor, underage, loli, shota negative stay in place, and any LoRA you use must be intended for adult subjects only. If a LoRA produces age-ambiguous output, lower its weight or stop using it.
Pairing build with checkpoint and style
Different checkpoints render the same build differently. Photoreal SDXL models give the most lifelike adult skin and proportions. Pony and Illustrious lean stylized, so curvy and athletic builds read in their respective art styles. Flux handles nuanced, descriptive builds well when you write full sentences. Our best checkpoints roundup and art style prompts help you match a build to the right base model and visual style.
For photoreal adult work specifically, our realistic results guide shows how build descriptors combine with skin texture and lighting to read as a real adult human rather than a render.
Build descriptors and proportions
It helps to understand what each descriptor actually changes so you can combine them intentionally. Build descriptors mostly affect three things: overall mass, the waist to hip ratio, and muscle definition. Slim and slender reduce mass and lengthen the silhouette. Athletic and toned add muscle definition without much mass. Curvy and hourglass push the waist to hip contrast. Plus size and full figured add mass and soften the silhouette. Muscular adds both mass and definition.
When you combine descriptors, make sure they describe complementary axes rather than fighting on the same axis. athletic plus curvy works because one controls definition and the other controls the waist to hip ratio. slim plus plus size does not work because both control mass in opposite directions, and the model will pick one or blend them unpredictably. Keeping descriptors on separate axes is the secret to combining them cleanly.
Proportions also depend on the body of the prompt around them. A close framing emphasizes the upper body and face, so build cues for shoulders and arms matter more. A full body framing shows the whole silhouette, so waist, hips, and legs descriptors carry more weight. Match your build descriptors to your framing so the most visible parts of the figure get described. Our camera angle prompts cover framing in detail.
Tasteful framing and outfits
Body type prompting and outfit prompting reinforce each other. The right clothing communicates build as clearly as a build descriptor, and it keeps the work tasteful. A fitted dress reads curvy, athletic wear reads toned, and a tailored suit reads broad shouldered. You can let outfits do some of the build work, which often produces a more natural, photographed look than stacking raw build tags. Our outfit prompts library pairs directly with this guide.
The same applies to setting and pose. A confident standing pose flatters an athletic build, while a relaxed seated pose suits a soft or curvy build. Our pose prompts and setting prompts help you choose a pose and environment that complements the build you want, all while keeping the character an unmistakable adult.
Putting it together
Good body type prompting is a short, disciplined formula: lead with a clear adult anchor and explicit adult age, add one primary build descriptor, add at most two supporting features, and carry the baseline safety negative on every render. That structure gives you intentional, repeatable adult characters across slim, athletic, curvy, plus size, and muscular builds.
The safety frame is not separate from the craft, it is part of it. Adults only, fictional AI characters only, never a real identifiable person, never a minor or minor-appearing subject, and child, minor, underage, loli, shota in the negative every single time. Discard anything ambiguous. With that baseline locked in, build a small library of recipes you trust, then run your favorites through our generator and refine one descriptor at a time until each adult character looks exactly the way you intended.
Frequently asked questions
How do I make the model respect a specific body type?
Lead with a clear adult anchor and an explicit adult age, then add one primary build descriptor such as athletic or curvy, then at most two supporting features. The anchor gives the model a target to lock onto so it stops defaulting. If the build is still ignored, add modest attention weighting around 1.2 to the build phrase, but keep weights gentle to avoid distorting anatomy.
Why does every example insist the character is an adult?
Because adult AI work has a hard, non-negotiable rule: subjects must be adults, fictional, and AI-generated, never minors or minor-appearing. Explicit adult anchors and an adult age push the model toward adult proportions and reinforce the safety negative. This is responsible craft, and it also produces more consistent adult characters. Any render that looks age-ambiguous should be discarded and re-prompted immediately.
What words should I never use in a body type prompt?
Never use youthful, young, teen, schoolgirl, loli, shota, petite used as an age signal, or any phrasing that suggests a minor or age regression. Body descriptors are about adult build only. Keep child, minor, underage, loli, shota in the negative on every prompt. If petite is used at all, it refers strictly to adult frame size and must sit beside a clear adult age anchor.
How do I keep the same body type across several images?
Reuse the same seed to stabilize proportions, then add a reference based tool. IPAdapter carries a character’s look across images, and LoRAs or embeddings lock build even harder. For posed sets, OpenPose fixes the skeleton while you vary other attributes. Combining a stable seed, a strong build descriptor, and a reference tool gives the most reliable adult character continuity across a set.
Can I generate two characters with different builds in one image?
Yes, with regional prompting. A plain prompt tends to blend two builds into an average, so assign each build descriptor to a separate region of the canvas using a tool like Regional Prompter. Apply the adult anchor and the baseline safety negative to the whole prompt, and the individual build descriptors to each region so each adult character keeps a distinct, intended physique.
Does weighting a build descriptor override safety rules?
No. Weighting only changes how strongly a phrase is applied, not its meaning. It can never override the adult anchoring or the safety negative, which stay fixed on every render regardless of weights. Use modest weights between 1.1 and 1.3 to emphasize a build that is being ignored, and always keep the adult age, maturity word, and child, minor, underage, loli, shota negative in place.
Which checkpoint renders body types most realistically?
Photoreal SDXL based checkpoints give the most lifelike adult skin and proportions, which makes them the best choice for realistic builds. Pony and Illustrious render the same builds in stylized or anime forms. Flux handles nuanced, descriptive builds well when you write full sentences. Match the build to the base model using a checkpoint roundup, then refine with skin texture and lighting terms for added realism.
What do I do if a render looks ambiguous in age?
Discard it immediately and do not use it. Then strengthen the adult anchor by raising the stated age and adding a maturity word, add youthful and teen to the negative alongside the baseline child, minor, underage, loli, shota tokens, and regenerate. Age ambiguity is never acceptable in adult AI work, so re-prompting until the subject is unmistakably an adult is the only correct response.
