Not a photoshoot.
Not a portrait.
Not “AI girlfriend” bait.
I tried generating the version of an AI beauty that only someone close enough would ever see.
No makeup polish.
Bad phone-photo energy.
Still impossible to ignore.
No-reference prompt anyone can try below.
Show me your AI beauty’s off-camera side.
This was the failed version.
The face was very close to what I wanted, but the setup wasn’t physically clear enough. The camera, the sunlight, and her position didn’t quite agree with each other, so the image drifted toward a selfie-like shot.
Still… if this exact face had landed with the final camera angle and a truly bare face, I probably would’ve called it done lol.
Prompt:
A fictional adult 20-year-old Japanese woman in an off-duty everyday moment. Early morning in a dark apartment bedroom, all lights off, door closed. One narrow gap in blackout curtains outside the frame lets in direct morning sunlight. The window, curtains, and light gap are not visible; the sunlight is only seen through the bright patch on her skin, not as a visible beam.
She has just woken up and sits upright near the pillow, wearing a worn oversized T-shirt as sleepwear, messy flattened hair, and a completely bare face. A close handheld smartphone photo by someone nearby on the same bed. The framing is face-dominant, close and intimate, not a photoshoot.
She raises one hand near her face to soften the sunlight. The hand partially breaks the light before it reaches her eyes, creating uneven shadow across her brow and one eye. She looks sleepy, mildly disoriented, and unfocused, not posing.
Harsh sunlight clips slightly on the lit skin, while the dark side falls into heavy shadow with subtle digital noise. Visible pores, uneven skin tone, under-eye dullness, dry lips, fine facial hairs, imperfect morning skin texture, natural hand and wrist. Beautiful but unpolished. No glamour, heavy makeup, beauty-filter skin, cinematic lighting, professional posing, anime, text, or watermark.
Does she feel real, or does the AI still show?
I built this by cutting away the unnatural setup:
no makeup → just woke up → curtains still closed → one slit of morning light → she keeps one hand up against the morning light before fully waking up.
The goal was a low-light smartphone photo where the realism comes from the limits of the camera, the light, and the body’s reaction.
GPT-image-2 on ChatGPT
Prompt below ↓
The shoulder is a little off in this one, but I’m letting it slide — I’ve already done too many rerolls.
Here’s a shorter version of the prompt. It loses a bit of precision, but it should still get you close.
Prompt:
An unusually beautiful fictional adult 20-year-old Japanese woman in an off-duty everyday moment, captured as a casual handheld rear-smartphone photo, not a selfie. Early morning in a dark apartment bedroom, all lights off, door closed. A single narrow slit of low morning sunlight enters almost horizontally from the left through an unseen gap in blackout curtains outside the frame. The window, curtains, gap, sky, and exterior are not visible. The sunlight is seen only by its effect as one narrow hard-edged patch of light, not as a visible beam in the air.
She has just woken up and is sitting upright near the pillow, wearing a worn oversized T-shirt used as sleepwear, with a slightly stretched collar, messy sleep-flattened hair, and a completely bare face. A separate unseen photographer is seated opposite her at the near edge of the bed, about 80–100 cm away, using the rear camera of a smartphone from a natural face-to-face distance, held horizontally slightly below her eye level. The shot is a close bust-up portrait. Most of the bed, bedding, and room are cropped out or nearly invisible.
Her face turns toward the left-side sunlight, creating a left-facing three-quarter view close to profile. She keeps her left hand raised near her brow to soften the incoming light. Only the left hand, wrist, and a small part of the forearm are visible. The hand partly breaks the slit-shaped sunlight before it reaches her face, so the light falls as an irregular narrow hard-edged patch across the side of her hand, fingers, brow, cheekbone, nose edge, lips, and part of the lit side of her face.
She looks freshly awake, sleepy, and mildly disoriented, not posed. She does not look upward. Her head stays level, her chin remains neutral or slightly lowered, and her gaze is horizontal, unfocused, and not directed at the lens. The phone exposes for the harsh sunlit skin, so the brightest lit areas may clip slightly, while the shadow side falls into deep dark tones with subtle digital noise. Her visible skin should feel physically real and unpolished: visible pores, slight uneven skin tone, faint under-eye darkness, subtle redness, natural dry lips, mildly imperfect lived-in hair, tiny imperfections, and tactile morning texture. She should remain obviously beautiful, but without glamour, heavy makeup, whitening, perfect composition, cinematic lighting, polished hair, doll-like features, professional posing, pin-up posing, fetish framing, hidden-camera 느낌, anime style, text, or watermark.
@IamEmily2050 Decided to take my AI pfp to the raw camera casting treatment. Turns out she just needed to wash her face and ditch 10 layers of digital foundation 😂
Your prompt is insane — the realism hits different. Thanks!
The key tweak was defining two things more clearly:
1. her eyes are focused on her own reflection, not just looking “past the camera”
2. the redness is post-wash flush, not blush or leftover makeup
Prompt:
Generate 1 image.
Scene:
late at night in a small shared apartment bathroom just after fully washing off all makeup, standing close to the sink and looking at her own reflection rather than the lens, hair loosely tied back with a few damp strands stuck to her cheek, harsh fluorescent light overhead, photographed from beside her at the sink by someone close enough to stand shoulder-to-shoulder with her, smartphone held near cheek height and angled slightly across her face, very close portrait with only her face, neck, and a sliver of one shoulder visible, bare skin freshly cleansed and slightly towel-dried, temporary uneven post-wash redness concentrated around the nose, cheeks, and chin rather than blush-like color, no visible eyeliner, mascara, lip tint, blush, or foundation residue, close enough to see faint pore texture, dry lips, and tired under-eyes, her eyes softly focused on the mirror with small natural catchlights, slight moisture along the lower waterline, relaxed eyelids, and a quietly occupied, slightly tired expression rather than a blank or glassy stare.
Subject:
A fictional, clearly adult 20-year-old Japanese woman in an off-duty everyday moment.
Core Intent:
Start from an exceptionally beautiful AI-beauty archetype, then remove the performance.
She should be obviously beautiful at first glance, but the image should feel candid, intimate, and unusually lifelike rather than glamorous, polished, or posed.
Beauty Baseline:
She remains exceptionally beautiful, with a small balanced face, expressive eyes, refined features, and slim elegant proportions with a narrow waist and long slim legs.
Her beauty should read through the whole person, not only the face, and should remain clear even without flattering styling, makeup, lighting, or posing.
Realism:
Visible pores, slight uneven skin tone, faint under-eye darkness, subtle redness, natural lips, mildly imperfect lived-in hair, and small incidental imperfections.
The realism should feel tactile and alive, not uncanny, synthetic, or stylized.
No beauty-filter skin.
Grooming:
Private scenes may be fully makeup-free and relaxed.
Public scenes should use minimal, understated, socially plausible grooming.
Do not default to polished no-makeup makeup.
Camera / Framing:
Casual handheld smartphone snapshot by someone close and trusted.
Use a physically plausible handheld viewpoint: the camera should be in a place where that person could actually stand, sit, lean, or hold a phone in the scene.
Slightly awkward crop, slight tilt, mild blur, imperfect focus, uneven exposure, ordinary phone color.
Not a photoshoot.
Composition should follow the scene naturally: close, medium, or full-body framing are all acceptable.
Prefer natural mid-action moments over centered standing portraits.
Scene Adaptation:
Outfit, gaze, expression, posture, hairstyle, and grooming should adapt naturally to the setting.
Negative Prompt:
glamour, heavy makeup, whitening, perfect composition, cinematic lighting, polished hair, doll-like face, average face, beauty-filter skin, professional posing, pin-up posing, fetish framing, hidden camera, surveillance, anime, text, watermark.
I tried generating a close-up “just washed off makeup” face.
Two things kept bothering me:
the redness looked too much like makeup, and when the gaze was off-camera, the eyes sometimes looked dead.
Left: vague off-camera gaze + makeup-like redness
Right: mirror-focused gaze + post-wash skin texture
The fix was adding what the eyes are actually focused on, plus defining the redness as post-wash flush, not blush.
Looks a lot better, right?
No-Reference Prompt Anyone Can Try
prompt below ↓
Haha, now I'm confused which Grok to believe 😂Anyway, better safe than sorry — keeping the sensitive label on. That's probably the healthy way on Elon’s X anyway.
Haha exactly 😂 Wet hair clinging with those glistening droplets + the sheer fabric catching light and water spots is pure classifier catnip. It over-indexes on texture and "intimate lighting" cues even when it's clearly just a cozy, fully clothed bedroom moment.
Human eyes see tasteful realism. The algo sees a gray-area minefield. Classic over-sensitivity. Nice prompt work on the photoreal details!
@Adam38363368936 Love this. Keeping the AI beauty baseline while going fully off-duty — raw skin, lived-in hair, casual phone-snap vibe — creates the version only someone close would see. No filters, no performance, just real presence.
For this one, I only changed the Input part to push a warmer, more intimate summer-night mood.
Example input version prompt:
Generate 1 image.
Fictional 20-year-old Japanese woman.
Core:
Start from a polished AI-beauty archetype, then show her off-duty in everyday life. Keep elite face/body baseline: small balanced face, expressive eyes, refined features, slim elegant proportions, long slim legs, and a striking yet believable AI-beauty presence. Off-state must not make her plain.
Realism:
Visible pores, uneven tone, under-eye dullness, subtle redness, tiny imperfections, dry lips, lived-in hair, no beauty-filter skin.
Photo:
Casual handheld phone snapshot by someone close/trusted. Awkward crop, slight tilt/blur, uneven exposure, ordinary phone color. Not a photoshoot.
Context:
Private setting, so she can be almost makeup-free, with slightly messy natural hair and relaxed off-duty styling.
Input:
A hot summer night at home in the bedroom. She is relaxing on the bed in a very light, slightly sheer camisole slip-style loungewear dress, intimate and private but still natural, not staged. She is looking toward the camera with a soft inviting expression, as if wordlessly telling someone close “come here” or “don’t keep me waiting.” Her expression is gentle, warm, slightly sleepy, and subtly suggestive without exaggerated posing. Her hair is down and a little messy. The image should feel intimate through eye contact and mood, not through overt glamour.
Negative:
glamour, heavy makeup, whitening, perfect composition, doll face, plain face, fetish framing, hidden camera, surveillance, anime, text, watermark.
Try it. Only change the Input part.
The prompt keeps the AI-beauty baseline, then lets the scene decide how off-duty she becomes.
Input format:
[where she is] + [what she’s doing] + [gaze / expression] + [private or public mood]
Result:
an off-duty AI beauty — still unrealistically attractive, but with raw skin, imperfect hair, socially natural styling, and bad phone-photo energy.
Prompt summary:
Generate 1 image.
Fictional 20-year-old Japanese woman based on Input.
Core:
Start from a polished AI-beauty archetype, then show her off-duty in everyday life. Keep elite face/body baseline: small balanced face, expressive eyes, refined features, slim elegant proportions, long legs. Off-state must not make her plain.
Realism:
Pores, uneven tone, under-eye dullness, redness, imperfections, dry lips, lived-in hair, no beauty-filter skin.
Photo:
Casual handheld phone snapshot by someone close/trusted. Awkward crop, slight tilt/blur, uneven exposure, ordinary phone color. Not a photoshoot.
Context:
Grooming adapts to scene. Private can be makeup-free/messy; public uses minimal makeup and plausible hair. Quiet, discreet, no crowd.
Input:
[quiet train box seat, looking outside]
[relaxing on bed]
[camisole and shorts, by a fan]
Negative:
glamour, heavy makeup, whitening, perfect composition, doll face, plain face, fetish framing, hidden camera, surveillance, anime, text, watermark.
Not a photoshoot.
Not a portrait.
Not “AI girlfriend” bait.
I tried generating the version of an AI beauty that only someone close enough would ever see.
No makeup polish.
Bad phone-photo energy.
Still impossible to ignore.
No-reference prompt anyone can try below.
Show me your AI beauty’s off-camera side.
No-reference prompt anyone can try.
If you want to customize the result, the easiest things to tweak are:
- beauty level
- realism level
- setting privacy
- camera roughness
- 4-image variation
Full prompt:
Create 4 separate independent image outputs, not a collage. Generate 4 different fictional 20-year-old Japanese women, all different people.
Start from the familiar polished AI-generated Japanese beauty archetype, then show each woman in her off-state of everyday life. She must remain exceptionally beautiful, with top-tier natural facial harmony, even without styling or polish. Do not let the off-state make her plain or average.
Each image must feel like a separate slice of real life, with clearly different setting, framing, mood, body language, expression, gaze, and camera position. Avoid repeated templates.
Make the photography feel like casual handheld smartphone snapshots taken by someone very close to her and trusted by her, not by a photographer. No professional composition, no cinematic lighting, no beauty-shot polish. Allow awkward framing, casual cropping, uneven exposure, mild blur, slight tilt, ordinary phone color, and imperfect focus.
If a public or semi-public setting appears, it must feel quiet, discreet, and socially plausible, with no strangers immediately nearby and no crowded public scene. Hair, makeup, and styling must match the privacy level of the setting: private scenes can be makeup-free, while public scenes may have only minimal understated everyday makeup. Never glamorous, never over-whitened, never polished. Keep natural skin tone and socially natural hair.
Do not script specific actions. Capture each woman in an unperformed in-between moment. Raw skin realism: visible pores, slight uneven tone, faint under-eye dullness, subtle redness, tiny imperfections, natural dry lips, no beauty-filter smoothness. Natural imperfect hair with roots, loose strands, flyaways, uneven volume, lived-in texture.
Across the 4 images: 1 looking away, 1 with brief natural eye contact, 1 with a soft unguarded smile, 1 neutral or slightly tired.
Negative prompt: same person, same scene, same framing, same face angle, same gaze, same expression, professional photoshoot, perfect composition, cinematic lighting, glamour photography, heavy makeup, glossy skin, excessive whitening, polished hair, doll-like face, plain face, average-looking face, crowded train, adjacent strangers, voyeuristic angle, hidden camera, surveillance, anime, text, watermark.