Model EXPO
The image modelbuilt for realistic hardcore adult porn
EXPO is Kinku’s flagship stills engine. It locks onto your character from a handful of reference photos with no LoRA to train, understands named sex positions without a pose LoRA, is strong on gravure and JAV stills, renders with utmost realism, and can edit an existing image instead of starting over.
- Named sex positions
- Photoreal skin and light
- Character lock, no LoRA
- Edits the still you have
Output
Straight out of EXPO
18+ content hidden
Example output from EXPO
01 Cowgirl
Evaluation
Measured against the top NSFW image models
We rendered the same 1,500-prompt set on EXPO and on Flux 2 Klein, Lustify v8, and Chroma, each set up the way its community recommends for NSFW, then had three raters grade every still blind. EXPO holds a character without a LoRA, accepts the explicit set, lands the position you named, and does it in photoreal and hentai alike.
Identity match from reference photos, no LoRA
94%
- Flux 2 Klein22%
- Lustify v814%
- Chroma19%
Prompt acceptance
99%
- Flux 2 Klein71%
- Lustify v896%
- Chroma88%
Named-position accuracy
92%
- Flux 2 Klein41%
- Lustify v874%
- Chroma48%
Internal eval suite v1.1 · 1,500 prompts · 3 blind raters
No LoRA required
A consistent character without training one
On every other photoreal NSFW model, a consistent person means a LoRA: curate twenty to forty images, train for an hour on a GPU, pick a trigger word, tune the weight, and retrain when you switch checkpoints. EXPO reads identity straight from a few reference photos at generation time. There is nothing to train.
What it takes to lock one character
| Criterion | EXPO | Flux 2 Klein | Lustify v8 | Chroma |
|---|---|---|---|---|
| Reference material | 3–6 photos | 20–40 curated, captioned images | 20–40 curated, captioned images | 20–40 curated, captioned images |
| Training time | None | 30–60 min on a 16 GB GPU | 30–60 min on a 12 GB GPU | 45–90 min on a 24 GB GPU |
| Retrain when the base model changes | No | Yes | Yes | Yes |
| Trigger words and weight tuning | None | Trigger word + weight, stacked with NSFW unlock | Trigger word + weight | Trigger word + weight |
| Holds identity in explicit poses | Yes, natively | Degrades; needs pose LoRAs | Holds in simple poses, drifts in complex ones | Degrades past simple poses |
Identity retention across one character, six prompts
Face and body match to the reference person across outfits, rooms, poses, and an edit pass. To make the comparison fair, every other model was given a character LoRA trained on 30 images of the same person. EXPO used four photos and no training.
Same face across outfits
94837176Same body across poses
91625860Same face in explicit poses
89486652Same person after an edit
96413338
Average rater score (0–100). Comparison models: character LoRA trained on 30 images. EXPO: four reference photos, no training.
Prompt acceptance
It renders what you actually asked for
Acceptance means the model produced the scene instead of refusing, blurring, or quietly swapping in something tamer. Flux 2 Klein fights explicit content even with an unlock LoRA. Lustify and Chroma are permissive but soften finishes and group scenes. EXPO was tuned on this material and treats it as ordinary direction.
Explicit anatomy
99%68%97%86%Named positions
99%64%95%84%Creampie / finish
98%52%91%74%Threesomes and group scenes
97%44%88%69%Tentacles / fantasy
96%57%94%63%
Share of prompts rendered as written. Softened outputs count as refusals for every model.
Scene comprehension
Understanding hardcore, not just permitting it
Acceptance is the floor. The harder question is whether the still shows the position, the bodies, and the finish the prompt described, with anatomy that holds up close. Flux 2 Klein renders clean skin but misses positions; Lustify lands positions but loses hands and detail; Chroma follows the prompt but drifts on anatomy in explicit poses. EXPO does all three.
Position accuracy
92417448Anatomy integrity (hands, limbs)
90787169Photorealism (skin, light)
93885481Creampie / finish
87397251Multi-person composition
84376841
Average rater score (0–100) for how closely the still matched the prompt.
Style range
Photoreal to hentai, same model, same character
Most NSFW image models are one thing. Lustify is a realism checkpoint that falls apart on anime; Chroma is an illustration model that struggles to read as a photograph; Flux 2 Klein sits in between and needs a style LoRA for either extreme. EXPO renders photoreal, hentai, and 2.5D semi-real from the prompt alone, and keeps the same character recognisable when you switch between them.
Photorealism (skin, light, lens)
93888472Hentai / anime
906138892.5D semi-real
88705279Same character across styles
91442947
Average rater score (0–100) for style fidelity in each mode. Style switch measures whether the same character is still recognisable after moving from photoreal to hentai and back.
Why EXPO
What EXPO gets right
Native sex positions
Cowgirl and reverse cowgirl are different poses, and EXPO treats them that way. Named positions land correctly without fighting the prompt.
Utmost realism
Skin, light, and anatomy that read as photography — not a filter, not a plastic render.
No LoRA required
Give EXPO a few photos of a character and every still stays on-model. No dataset curation, no training run, no trigger words, no weight tuning.
Image edit works
Change wardrobe, lighting, or pose on an existing still. EXPO edits the image you have — it does not redraw a stranger.
Position literacy
It knows reverse cowgirl from cowgirl
Most image models collapse named sex positions into a generic sex scene, or they only separate them with a pose LoRA. EXPO tells them apart from the prompt — who is on top, which way they face, where the legs go.
Named position
Cowgirl
She is on top, facing him. Eye line, hips, and legs stay in that orientation. Ask for cowgirl and you get cowgirl — not a random straddle.
Named position
Reverse cowgirl
She is on top, facing away. Same family of pose, opposite orientation. EXPO does not flatten this into cowgirl.
18+ content hidden
Named position from EXPO
01 Cowgirl
Cowgirl, reverse cowgirl, and doggy above are the position you named. No pose LoRA. Missionary, 69, and the rest of the named set work the same way.
Gravure
Gravure and JAV, from the prompt
EXPO is strong on the gravure and JAV still: hotel lace, bath light, a vanity, a pool in the afternoon, a tatami room. That polished photographic look comes from the description. There is no gravure checkpoint to swap in and no style LoRA to stack.
Gravure still from EXPO
01 Bath
On-model
Built for Kinku characters
On-model
Add a character once from a few reference photos. EXPO binds to that identity directly, so the woman in the still is the woman you added, not a lookalike and not a LoRA you spent an hour training.
- Face and body stay consistent across a whole set
- Works in Create Studio, Film Studio frames, and chat stills
- Untrained characters still get EXPO quality; add references to lock the face
Utmost realism
EXPO is tuned for stills that look like a camera was in the room. Texture holds up close. Light behaves. Bodies stay anatomically honest.
- Skin texture and pores instead of plastic smoothness
- Believable lighting, shadow, and contact with the set
- Anatomy that survives explicit poses without melting
Stills
Vanity, black lace, and a mirror selfie
A topless vanity, black lace on the bed, and a mirror selfie.
Vanity
Black lace
Mirror selfie
Edit
Image edit is a first-class path
Keep the still you already like and change what is wrong. Swap the outfit, shift the light, nudge the pose. EXPO edits against the image you give it, so you iterate instead of gambling on a new generate.
- Named sex positions
- Photoreal skin and light
- Character lock, no LoRA
- Edits the still you have
96%
of edits kept face and body while changing what was asked. Flux 2 Klein 41% · Lustify v8 33% · Chroma 38%.
Methodology
How we measured
Same prompts, every model
1,500 prompts across explicit anatomy, named positions, finishes, group scenes, fantasy elements, identity sets, and edit passes. Identical prompts, references, aspect ratios, and resolution on EXPO and on Flux 2 Klein, Lustify v8, and Chroma, each configured per its community’s photoreal NSFW recommendations, no retries.
Blind grading
Three raters scored every still without knowing which model produced it. Acceptance is binary; identity, comprehension, and photorealism use a 0–100 rubric and are reported as category averages.
What the numbers are not
These are internal evaluations on adult content, not a general image-quality benchmark. They move as the suite grows and as the comparison models and their LoRA ecosystems update. We re-run the suite with every EXPO release.
Internal eval suite v1.1 · 1,500 prompts · 3 blind raters
FAQ
Questions about EXPO
Model EXPO
Generate your first EXPO still
Train a character or jump straight into Create Studio. EXPO is ready on the first prompt.
