What Is Prompt Seen? Copy-Paste AI Photo Prompts Guide

If you have scrolled Instagram, Reels, or TikTok this year and wondered how ordinary people are posting magazine-grade portraits of themselves, the answer is almost always the same two words. Prompt Seen is the search term people use to find the written instructions behind viral AI photos — and this guide explains exactly what it is, why it took over in 2026, and how to get the same results in about ninety seconds.

Quick answer: Prompt Seen is not an app, a filter, or a filter pack. It is the name of a copy-and-paste workflow: you find the text prompt behind a viral AI photo, paste it into ChatGPT or Gemini along with a clear photo of yourself, and the model applies that style while keeping your real face. The phrase spread because people literally searched for the prompt they had seen.

What "Prompt Seen" actually means

The confusion is understandable, because "Prompt Seen" sounds like a product. It is not.

It is not a filter. There is no "Prompt Seen" effect you can tap. Every result is generated fresh by a model from a written instruction.
It is not one model. The same prompt produces different results in Gemini, ChatGPT, and Grok.
It is a library format. A Prompt Seen page follows one pattern: show an example image, publish the exact prompt in plain text, let the reader copy it, upload their own photo, and generate their own version.

The name traces back to early prompt-sharing sites and apps built for Indian creators, where the phrase became shorthand for "the prompt from that photo I saw." It has since spilled into global English usage, which is why the same term now shows up attached to dozens of style variations.

You will see it written several ways

There is no official spelling. People write Prompt Seen, PromptSeen as one word, or even the typo "promote seen" — all of them lead to the same idea, so if you arrived here through one of those, you are in the right place.

The more useful distinction is what comes after the name, because that tells you what to expect: a 3D Prompt Seen turns a photo into a figurine, Prompt Seen boy or Prompt Seen girl narrows the look by who is in it, and Prompt Seen ChatGPT or Prompt Seen Gemini filters by the tool you plan to use.

Why it exploded in 2026

Two things had to line up, and in 2026 they finally did.

1. Models learned to keep the face. This is the single biggest factor. Earlier image tools could restyle an image, but they also replaced the person. Identity preservation only became reliable with the current generation of image models — Gemini's Nano Banana line, ChatGPT's image generation, and Grok's image tools. Once your own face survived the edit, the whole thing stopped feeling like a gimmick and started feeling like a photo shoot.

2. The before-and-after format is native to short video. A fifteen-second Reel that shows an ordinary selfie on the left and a cinematic portrait on the right is inherently watchable, and the question it provokes is always the same one in the comments: prompt? That comment is the demand signal, and it is what sends people to a search box — where whole sites of copy-ready prompts are waiting to answer it.

The catch is the lifecycle. Individual styles burn out quickly. A look that dominates Reels this week can feel dated within a month once enough people have posted their version. So the useful skill is not memorising one prompt — it is understanding the structure that makes any of them work.

How Prompt Seen works: three steps

Pick a look and get the prompt. Choose the style you want — cinematic portrait, 3D figurine, retro film, cartoon — and copy the full prompt text. It must be the complete prompt, including the identity-preservation line, not a summary of it.
Upload a clear photo of yourself. Front-facing, good daylight, no sunglasses, no group shots, and your face reasonably large in the frame. This matters more than the prompt does.
Paste, generate, then correct. Send the prompt with the photo attached. If the result drifts — wrong outfit, wrong lighting, a face that is almost yours — do not rewrite the whole prompt. Send a single correction line instead: keep my face exactly the same, only change the outfit.

Which tool should you use?

Tool	Image model	Best for	Watch out for
Gemini	Nano Banana	The most accurate face preservation and the most predictable prompt following	Occasionally needs a one-line correction to stop the style from overwriting the face
ChatGPT	ChatGPT image generation	Natural skin tones and believable textures; strong with couples and two-person scenes	Slower, and it more often refuses or softens realistic edits of real people
Grok	Grok image tools	Fast experimentation and trend-style looks	Less consistent identity preservation across repeated generations

If a tool declines a realistic edit of a real person, add the phrase as a stylised portrait to your prompt and resend. That reframing resolves the majority of refusals.

The anatomy of a working prompt

This is the part most prompt lists skip, and it is the reason two prompts that look similar can produce wildly different results. Every reliably working Prompt Seen prompt has five parts, in this order:

1. The identity lock. One sentence, always first: Use the uploaded photo as reference. Keep the exact same face, skin tone, and facial features. Do not change my identity. This is the line that separates a Prompt Seen edit from a random AI person. Moving it to the middle of the prompt measurably weakens its effect — put it first.

2. The styling instruction. Wardrobe, hair, and accessories, stated concretely: cream oversized shirt, minimal gold hoops, hair loosely curled. Vague words like "stylish" or "aesthetic" give the model nothing to aim at.

3. The lighting direction. Name the source and the quality: low sun behind the subject creating warm rim light and a soft lens flare, or single soft key light from the upper left with a subtle hair light. Lighting does more for perceived quality than any other single element.

4. The camera and composition. Lens, depth of field, framing, and aspect ratio: shot on 85mm, shallow depth of field, portrait 4:5, centred composition. This is what makes the output read as photography rather than illustration.

5. The output specification and guard lines. Resolution, realism markers, and the things you do not want: ultra-realistic, natural skin texture, visible pores, no text, no logo, no watermark, no distorted hands.

A prompt missing parts 1 and 5 will still generate something. It just will not generate you.

Twelve prompts you can copy right now

Every prompt below follows the five-part structure. Replace the bracketed details with your own, keep the identity lock exactly as written, and adjust one variable at a time so you can tell what changed the result.

1. Prompt Seen girl — soft cinematic portrait

Best for profile pictures and everyday posting.

Use the uploaded photo of an adult woman as the identity reference. Keep the exact same face, skin tone, hairstyle, and facial features. Do not change my identity. Create a soft cinematic portrait: cream oversized shirt, delicate gold earrings, hair with slight natural movement. Place her beside a large sunlit window, warm diffused daylight, gentle shadows, shallow depth of field. Shot on 85mm, natural skin texture, no heavy retouching, subtle film grain. Ultra-realistic, 4K, portrait 4:5, no text, no watermark, no distorted hands.

Tweak: swap the window for a plain studio backdrop to make it read as an editorial headshot instead of a lifestyle photo.

2. Prompt Seen boy — street portrait at golden hour

Best for the "trending boy" profile edit.

Use the uploaded photo of an adult man as the identity reference. Keep the exact same face, skin tone, hairstyle, and facial features. Do not change my identity. Create a cinematic street portrait: plain overshirt over a t-shirt, quiet city street at golden hour, warm rim light from behind, softly blurred background, relaxed confident expression, natural fabric texture. Shot on 50mm, shallow depth of field, subtle film grain. Ultra-realistic, 4K, portrait 4:5, no text, no logos, no watermark.

Tweak: shift to overcast light and grey tones for a cooler, moodier version of the same shot.

3. Prompt Seen attitude boy — low-angle editorial

Best for a bolder, more confident feed post.

Use the uploaded photo of an adult man as the identity reference. Keep the exact same face, skin tone, and facial features. Do not change my identity. Create a low-angle editorial portrait: dark fitted jacket, hands relaxed, direct gaze into the lens, clean concrete wall behind him, hard side light creating strong shadow gradients. Wide-angle perspective, high contrast, crisp clothing detail, film grain. Ultra-realistic, 4K, portrait 4:5, no text, no watermark, no distorted anatomy.

Tweak: raise the camera angle back to eye level if the low angle distorts the face.

4. Prompt Seen couple — golden-hour rooftop

Best for anniversaries and couple posts. Works from two uploaded reference photos.

Use both uploaded photos of the same two adults as the identity reference. Keep both faces, skin tones, hairstyles, and natural proportions exactly the same. Do not change either identity. Create a warm golden-hour rooftop portrait: the two standing close together, relaxed expressions, softly blurred city skyline, warm sunset rim light, realistic shadows, natural interaction that stays subtle and affectionate. Shot on 85mm, shallow depth of field, subtle film grain, premium lifestyle photography. Adults only, ultra-realistic, 4K, portrait 4:5, no text, no logos, no watermark, no duplicated limbs.

Tweak: if the two faces start to blend, generate each person separately first and describe the second person's outfit in text.

5. 3D Prompt Seen — collectible figurine

Currently the most durable style in the trend, and the one that ages least.

Use the uploaded photo as the identity reference. Keep the exact same face and likeness. Create a 1/7 scale commercialised PVC figurine of the person in the photo, realistic style, placed on a computer desk on a round transparent acrylic base. Behind the figurine, show the product packaging box with the character printed on it in official product art style. Soft indoor lighting, shallow depth of field, clean studio product photography, glossy plastic material, subtle reflections. Ultra-realistic, 4K, 4:5, no text, no watermark.

Tweak: ask for an action figure, a vinyl toy, or a marble statue instead of a PVC figurine for a different material feel.

6. Prompt Seen cartoon — stylised character

Best for playful avatars. Note that this one loosens the identity lock deliberately.

Use the uploaded photo as the identity reference. Keep the person's recognisable features, hairstyle, and overall likeness in a stylised form. Transform the subject into a high-quality 3D animated character in the style of a modern animated feature film: large expressive eyes, soft rounded facial features, simplified clothing with clean shapes. Warm cinematic lighting, soft shadows, minimal background, vibrant but controlled colour. Rendered look, 4K, square 1:1, no text, no watermark.

Tweak: reduce the stylisation by adding keep realistic facial proportions and natural skin texture to the prompt.

7. Prompt Seen retro — 1980s film photograph

Best for nostalgia posts and creative feed variety.

Use the uploaded photo as the identity reference. Keep the exact same face, skin tone, and features. Do not change my identity. Recreate the photo as an authentic 1980s film photograph: period-correct wardrobe, wide-collar shirt or patterned knitwear, vintage sunglasses optional. Setting of old streets, vintage shopfronts, classic architecture, or a parked period motorcycle. Warm analog lighting, realistic shadows, Kodak-style colour, faded tones, visible 35mm grain, natural photographic imperfections. Should look like a real photograph taken by a professional in the 1980s, never like a modern image with a retro filter. Vertical 4:5, no text, no watermark.

Tweak: change 1980s to 1990s and swap the wardrobe notes; the lighting and grain instructions stay identical.

8. Prompt Seen cinematic DP — moody profile portrait

Best for a strong WhatsApp or social profile picture.

Use the uploaded photo of an adult as the identity reference. Keep the exact same face, skin tone, and features. Do not change my identity. Create a cinematic profile portrait: dark turtleneck, hair neat and simple, dark moody background, dramatic side lighting with a subtle rim light separating the subject from the background, teal-and-orange colour grade, sharp catchlights in the eyes, visible skin texture. Shot on 85mm, shallow depth of field, film still quality. Ultra-realistic, 4K, square 1:1, no text, no watermark.

Tweak: square 1:1 is required for most profile pictures — do not leave it out.

9. Prompt Seen photo — warm aesthetic feed post

Best for travel and lifestyle content.

Use the uploaded photo as the identity reference. Keep the exact same face, skin tone, and features. Do not change my identity. Enhance this photo with a warm golden-hour glow: soft backlight, subtle lens flare, natural skin texture, relaxed expression, airy filmic colour grade, gentle highlight roll-off. Keep the composition close to the original. Ultra-realistic, 4K, portrait 4:5, no text, no watermark.

Tweak: avoid overexposing the face — if highlights blow out, add preserve facial detail, do not blow out the highlights.

10. Prompt Seen lofi — dusk filter

A newer Reels-driven look: low contrast, purple-blue haze, warm streetlight accents.

Use the uploaded photo as the identity reference. Keep the exact same face and features. Apply a lofi dusk aesthetic: muted purple-blue twilight tones, soft atmospheric haze, subtle film grain, low-contrast dreamy grade, warm streetlight glow in the background, nostalgic and quiet mood. Keep the subject's clothing and pose unchanged. Ultra-realistic, 4K, portrait 4:5, no text, no watermark.

Tweak: for an urban night version, add deep blue-violet shadows, distant neon bokeh, soft mist, desaturated palette with warm highlight accents.

11. Prompt Seen horror — moody dark portrait

Popular seasonally around Halloween. Preview the output before posting it publicly.

Use the uploaded photo of an adult as the identity reference. Keep the exact same face and features. Do not change my identity. Create a dark cinematic portrait: dim abandoned interior, single cold light source from below and behind, deep shadows across part of the face, cool desaturated grade, faint mist, unsettling but non-graphic atmosphere. Shot on 35mm, grain, high contrast. Ultra-realistic, no gore, no injury, 4K, portrait 4:5, no text, no watermark.

Tweak: keep the mood atmospheric rather than graphic; explore lighting and setting instead of disturbing content.

12. Festival and traditional looks

Adjust the outfit and setting to the occasion rather than reusing a generic prompt.

Use the uploaded photo of an adult as the identity reference. Keep the exact same face, skin tone, hairstyle, and features. Do not change my identity. Create an elegant traditional portrait: refined outfit with rich fabric texture and visible detail, tasteful jewellery, softly lit heritage interior with warm window light, natural posture, dignified styling, cinematic portrait photography, warm golden tones, shallow depth of field. Ultra-realistic, 4K, portrait 4:5, no text, no watermark, no distorted jewellery or hands.

Tweak: name the specific occasion and outfit so the model has a concrete target — traditional wear, festive wear, or formal wear all produce different results.

Before you paste: the photo matters more than the prompt

The single biggest cause of a disappointing result is the input photo, not the prompt. In order of importance:

Face size. Your face should fill a good part of the frame. Full-body shots far away from the camera give the model too little identity information.
Light. Daylight, evenly lit, no harsh shadow across the face. Backlit photos drift the worst.
Angle. Roughly facing the camera. Extreme side profiles and heavy tilts are ambiguous.
Clarity. Blurry or dark source photos reliably produce someone who is not quite you.
Occlusion. No sunglasses, no masks, no hands over the face, and ideally no other people in frame.

If the output looks like a stranger, do not rewrite the prompt. Upload a better photo.

Troubleshooting: six common failures

Problem	Likely cause	Fix
The face is not quite mine	Weak identity lock, or the photo is unclear	Move the identity lock to the very start of the prompt; upload a clearer, closer photo
The style took over the face	The model prioritised the styling instruction	Send: keep my face exactly the same, only change the outfit and lighting
The tool refused to generate	Realistic edits of real people trigger extra caution	Add as a stylised portrait and resend; reduce realism claims
Distorted hands or extra fingers	Long prompts with many objects and poses	Simplify the pose description; add no distorted hands, natural anatomy
The result looks like plastic	Over-retouching and missing texture cues	Add natural skin texture, visible pores, no heavy retouching, subtle film grain
It worked last month and now it does not	Trend drift or a model update changed behaviour	Rebuild from the five-part structure rather than reusing the old prompt text

Using Prompt Seen responsibly

This workflow is built around editing your own photo, and the etiquette is simple:

Only upload photos of yourself. Do not generate realistic images of other people without their permission.
Do not create intimate, compromising, or humiliating images of anyone, including yourself.
Never apply this to images of minors.
Avoid celebrity likenesses and impersonation. Do not generate images that suggest a real person said, did, or endorsed something they did not.
Label AI-generated portraits where your audience might reasonably assume a photo is real. A short caption is enough.

Most platforms restrict or remove synthetic images of real people, so staying inside these lines protects your account as well as the people in your photos.

Does Prompt Seen still work in 2026, and what happens next

The workflow is stable. The individual styles are not. Every few weeks a look reaches saturation — it has appeared on enough feeds that nobody stops scrolling for it — and the trend moves on to a new lighting treatment, a new material, or a new format.

That has one practical consequence for you: saved prompts go stale. A list published six months ago often contains styles that no longer read as current, and a prompt that worked beautifully in spring can fall flat by autumn. Check when a prompt list was last updated before you commit to it — and if a style feels overexposed on your feed, that is your cue to move on to the next one.

What is unlikely to change is the underlying skill: identity lock first, concrete styling, named lighting, camera direction, and an output specification. Those five parts will keep working long after today's specific looks stop being interesting.

Frequently asked questions

Is Prompt Seen an app? No. It is a copy-and-paste workflow — a text prompt plus a photo — that you run inside whichever image tool you already use. Some sites have adopted the name, but there is nothing to install.

Is it free? The prompts themselves are free to copy, and the free tiers of Gemini and ChatGPT are enough to generate results. Paid tiers mainly buy speed and higher limits, not quality.

Which is better, Gemini or ChatGPT? Gemini's image model preserves faces more accurately and follows detailed prompts more predictably. ChatGPT produces slightly more natural skin tones but is slower and more likely to refuse a realistic edit of a real person. Most people end up using both.

Why does my result look nothing like me? Almost always the source photo. Upload a closer, better-lit, front-facing image before touching the prompt.

Can I do this with a couple photo? Yes. Upload both photos and use the identity-lock line for both faces. Two-person generations drift more often, so expect to correct once or twice.

Do I need to know anything about prompt writing? No. That is the entire appeal — you copy the instruction, attach your photo, and correct with one line if something drifts.

Keep exploring

Once you have the structure down, the styles are just variations on it:

Prompt Seen boy — street, editorial, and attitude looks for the current feed
Prompt Seen girl — soft cinematic, retro, and studio portraits
Prompt Seen couple — anniversary, rooftop, and engagement scenes
3D Prompt Seen — figurines, action figures, and vinyl-toy versions
Prompt Seen cartoon — stylised animated-character portraits
Prompt Seen ChatGPT — prompts written and tested specifically for ChatGPT's image tools

Start with the structure, copy one prompt, and adjust a single variable at a time. That is the whole method — and it is why two words became a search term.