12 AI Photo Prompts for Instagram DP and Profile Pictures
A profile picture is the most viewed image on any Instagram account. It sits at the top of the profile, next to every comment, inside every...
Discover 14 best Gemini AI photo prompts for girls and boys in 2026. Create realistic, stylish portraits with trending ideas, prompt tips, and easy fixes.
Scroll far enough through any Indian feed in 2026 and the same shot keeps reappearing in variations: a familiar face, unmistakably real, standing in a palace courtyard at golden hour or leaning against a pillar on an empty metro platform after midnight. No photographer was booked. No location was paid for. The entire production was one selfie and a paragraph of text.
That paragraph is where everything is won or lost. Gemini will accept a three word instruction and return something generic, over smoothed, and faintly not quite the right person. The prompts that survive contact with a real feed read less like search queries and more like a shot brief handed to a photographer who has never met the subject.
Fourteen of those briefs follow. Five suit female subjects, five suit male subjects, and four work for anyone. All of them are written for adult subjects working from their own photograph, and each one closes with a short note on why it holds together plus the single detail worth changing first.
Most prompt lists still circulating were written for a model that no longer runs by default. The image side of Gemini now spans a small family of models rather than one, and knowing which one is answering explains a great deal about the results.
| Model | Technical name | Arrived | Where it fits |
| Nano Banana | Gemini 2.5 Flash Image | August 2025 | The original viral model. Still available, no longer the recommended default. |
| Nano Banana Pro | Gemini 3 Pro Image | November 2025 | Reasoning driven flagship. Up to 4K output and the strongest text rendering. |
| Nano Banana 2 | Gemini 3.1 Flash Image | February 2026 | Pro level quality at Flash speed. The default across the Gemini app. |
| Nano Banana 2 Flash Lite | Gemini 3.1 Flash Lite Image | 2026 | The lightest tier, aimed at high volume generation. |
Nano Banana 2 now replaces Nano Banana Pro across the Fast, Thinking and Pro modes in the Gemini app, while Google AI Pro and Ultra subscribers retain direct access to Nano Banana Pro. For portrait work the practical differences are worth knowing. Nano Banana Pro holds a recognisable likeness across as many as five people in a single frame, blends up to fourteen separate objects, and accepts six high fidelity reference shots. It generates at 2K and 4K where the original model capped at 1024 pixels. Through the API it lists at roughly $0.139 for a 2K image and $0.24 for 4K.
The second shift is quieter but matters more for prompt writing. Google's own guidance is direct about it: describe the scene rather than listing keywords. The model reads for meaning, not for tags. A narrative paragraph that names a time of day, a fabric, a lens and a mood will beat a comma separated pile of adjectives almost every time. Two related habits come from the same guidance. Negatives work better stated positively, so "an empty platform with no other passengers" outperforms "no people". And photographic vocabulary genuinely steers composition, which is why terms like 85mm, f/1.8, low angle and three point softbox appear throughout the prompts below.
Every prompt in this article is built on the same seven layers, in the same order. The order matters because the model resolves identity first and decoration last. Skipping the first layer is the single most common reason a face comes back looking like a cousin rather than the subject.
Layer 1: Identity lock. Name the uploaded image as the only facial reference and list what must not change. Explicitly forbidding beautification is the part most people leave out, and it is the part that keeps pores, asymmetry and real skin texture intact.
Layer 2: Wardrobe and styling. Name the fabric, not just the garment. Velvet, chiffon, raw silk, washed cotton and brushed wool each carry light differently, and the model knows the difference.
Layer 3: Environment. One specific place beats three vague ones. A sandstone haveli courtyard resolves cleanly. An "Indian background" does not.
Layer 4: Pose and expression. Give the body an angle and the hands a job. Unemployed hands are where AI portraits still fall apart.
Layer 5: Light. State direction, quality and colour. Warm sunlight from camera left with soft fill on the shadow side is a lighting plan. Cinematic lighting is a wish.
Layer 6: Camera. Body, focal length and aperture. This is the fastest route to real depth of field instead of a fake blur ring.
Layer 7: Grade and exclusions. Close with the colour palette, the aspect ratio, and a short list of what to leave out.
Use the uploaded photo of an adult woman as the only facial reference. Preserve her exact face shape, eye shape, brows, nose, lips, jawline, skin tone, age and natural expression, including visible pores and slight asymmetry. Do not smooth, slim or beautify the face. Dress her in a soft coral silk lehenga with antique gold zari embroidery and a sheer dupatta falling from one shoulder, the silk showing real weight and fold. Place her in the sandstone courtyard of an old haveli during the final twenty minutes of daylight, carved arches receding into warm haze behind her. Turn her body three quarters toward the camera with one hand resting lightly on a stone pillar, chin level, gaze steady and composed. Light the scene with low amber sunlight from camera left and a soft bounce filling the shadow side of her face. Photograph as if on a Canon EOS R5 with an 85mm lens at f/1.8, eyes tack sharp, the courtyard dissolving into amber bokeh. Grade toward warm sandstone, antique gold and deep shadow while keeping skin colour true. Vertical 3:4. No text, watermark, logo or distorted hands.
Three concrete anchors carry this one: a named window of daylight, a named fabric, and a named lens. Vague versions of the same idea produce a flat orange wash instead of directional light. First thing to change: swap the coral silk and the haveli for ivory chikankari and a Lucknow terrace, and the entire mood shifts from regal to quiet.
Treat the uploaded image of an adult woman as the definitive identity source and reproduce her real facial appearance with high fidelity, keeping bone structure, complexion, age and natural expression unchanged. Avoid any retouching that idealises her features. Drape her in a dove grey chiffon saree with a narrow silver border and a fitted three quarter sleeve blouse, the chiffon light enough to move. Seat her on a low cane stool beside a tall window in a plain whitewashed room, a single brass vase out of focus behind her. Angle her torso toward the window with her face turning back toward the lens, one hand resting on her knee and the other holding the pallu loosely. Illuminate her entirely with soft indirect daylight from that window, letting the far side of her face fall into gentle shadow with no artificial fill. Photograph as if on a Sony A7 IV with a 50mm lens at f/2, natural skin texture fully visible. Grade with muted pastels, cool greys and warm skin, low contrast and no heavy saturation. Vertical 4:5. No text, watermark or over smoothed skin.
Window light is the most forgiving lighting in portraiture and the least often requested. Removing artificial fill on purpose is what gives this look its softness rather than its flatness. First thing to change: move the window to a rainy afternoon and the same prompt turns melancholy without touching anything else.
Use the uploaded adult woman's photo as the sole face reference and keep her identity exactly intact, including natural skin texture and expression. Do not airbrush. Style her in a black satin slip dress with fine silver jewellery and glossy lips, hair slightly undone as if the evening has been going a while. Place her outside a restaurant on a busy street at night, warm signage and passing headlights blurred well behind her. Catch her mid turn toward the camera with a half smile, one hand adjusting an earring, body language casual rather than posed. Light her with a hard direct on camera flash that produces bright falloff, a crisp shadow behind her shoulder and slightly blown highlights on the cheekbones. Photograph as if on a compact point and shoot at 28mm, f/4, a touch of motion in the background. Grade with cool nighttime shadows, punchy saturation and the slightly harsh colour of early digital cameras. Vertical 4:5. No text or watermark.
The polished golden hour look has been running for years, and feeds have moved on. The current aesthetic favours harder flash, sharper shadows and imperfect exposure, which is exactly what this prompt asks for. First thing to change: requesting a mirror instead of a street produces the mirror selfie version, currently one of the most copied formats going.
Use the uploaded photograph of an adult woman as the exclusive facial reference, preserving her exact features, complexion, age and natural expression without beautification. Dress her in an oversized cream knit sweater with sleeves pushed to the forearm and small gold hoops. Seat her at a corner table inside an old cafe, a large rain streaked window beside her and the street outside reduced to amber bokeh. Have her cradle a cup in both hands, shoulders relaxed, looking out through the glass rather than at the lens, with a faint unforced smile. Light her with cool grey daylight through the wet window on one side and the warm glow of an interior lamp on the other, letting the two colours meet across her face. Photograph as if on a Fujifilm X-T5 with a 35mm lens at f/1.4, condensation and water trails on the glass rendered in focus near the edge of frame. Grade with teal shadows, amber highlights and gentle film grain. Vertical 3:4. No text, watermark or plastic looking skin.
Mixed colour temperature is an underused trick. Cool light on one side and warm on the other creates dimension that a single light source never delivers. First thing to change: replacing the cafe with a moving train window keeps the mood and adds a travel narrative.
Use the uploaded image of an adult woman as the single authoritative identity reference. Retain her exact facial geometry, complexion, age, expression and real skin characteristics, adding no digital perfection. Dress her in a deep emerald raw silk saree with antique gold zari work and a structured blouse with elbow length sleeves, the fabric substantial and clearly hand finished. Place her against a textured charcoal studio backdrop with nothing else in frame. Position her standing at a three quarter angle, one hand resting near the waist and the other holding the pallu, weight on the back foot, expression composed and quietly confident. Light her with a large softbox as key light from slightly above and camera right, a subtle rim light separating her hair and shoulders from the dark background, and controlled shadow beneath the jaw. Photograph as if on a Nikon Z8 with a 105mm lens at f/2.8, exceptionally sharp facial detail and strong subject separation. Grade with emerald, antique gold and near black, high dynamic range and no crushed shadows. Vertical 3:4. No text, logo, watermark or added props.
The empty backdrop is doing real work here. Removing the environment forces every pixel of attention onto wardrobe, light and expression, which is precisely how magazine covers are shot. First thing to change: switching the key light to a hard single source with no fill produces a dramatic chiaroscuro version of the same setup.
Use the uploaded photo of an adult man as the only face reference. Keep his facial structure, eyes, brows, nose, lips, skin tone, stubble and natural expression exactly as they are, with no smoothing or reshaping. Dress him in a tailored black bomber jacket over a plain white crew neck tee, dark straight leg trousers and clean minimal sneakers. Place him on an underground metro platform late at night, the platform empty of other passengers, tiled walls and cold overhead strip lighting stretching into the distance behind him. Have him lean lightly against a concrete pillar, one hand in his jacket pocket, shoulders squared, looking just past the lens with a calm unsmiling expression. Light him with the greenish overhead platform lights above and a faint cool spill from the tunnel, letting shadow gather under the brow. Photograph as if on a Sony A7 IV with a 35mm lens at f/1.8 from a slightly low angle, the platform receding into shallow focus. Grade with cool teal, concrete grey and desaturated skin, cinematic contrast, moderate film grain. Vertical 3:4. No text or watermark.
The low angle is not decoration. Shooting from slightly below adds presence, which is why the composition keeps recurring in male portrait trends this year. First thing to change: move the same setup to an empty multi storey car park and the mood turns colder still.
Use the uploaded adult man's photograph as the sole identity reference and preserve every recognisable feature, including skin texture, without retouching. Dress him in an unbuttoned cream linen shirt with the sleeves rolled, relaxed beige trousers and no visible branding, the linen showing natural creasing. Place him on a stone sea wall on a quiet coastline in late afternoon, the water behind him catching low light and a few boats out of focus on the horizon. Have him sit with one knee up and a forearm resting across it, looking out toward the water with a relaxed half smile, hair moving slightly in the wind. Light the scene with warm low sun from behind camera right, a strong rim of light along his shoulder and jaw, and soft natural fill bouncing off the pale stone. Photograph as if on a Canon EOS R6 with a 50mm lens at f/2, sharp on the eyes with the sea softened behind. Grade with warm sand, faded blue and sunlit skin, gentle contrast and a light haze in the highlights. Horizontal 4:3. No text, watermark or over saturated colour.
Understated wardrobe reads as expensive far more reliably than logos do. Linen, real creasing and an absence of branding are what sell this one. First thing to change: trade the sea wall for a vineyard terrace and the palette moves from coastal to Mediterranean.
Use the uploaded photo of an adult man as the definitive facial reference, keeping his exact features, skin tone and expression unchanged with no beautification. Dress him in a black leather jacket over a dark hooded sweatshirt, dark jeans and heavy boots. Place him in a narrow city alley after rain at night, wet asphalt reflecting blue and magenta signage, steam rising from a vent behind him. Have him stand facing the camera with his weight on one leg, hands loose at his sides, jaw set, expression direct and unbothered. Light him with coloured neon from both sides, magenta from camera left and cold blue from camera right, with almost no fill so the centre of his face falls into shadow. Photograph as if on an ARRI style cinema look, 40mm, f/2, shallow depth with the alley collapsing into coloured bokeh. Grade with deep blacks, saturated magenta and cyan, high contrast and visible grain. Vertical 3:4. No text, watermark or unreadable signage.
Two coloured sources from opposing sides is a lighting setup, not a filter, and the model responds to it as one. Asking for almost no fill is what keeps the result cinematic rather than flat. First thing to change: switch magenta and cyan for amber and deep green and the reference moves from cyber noir to something closer to a crime thriller.
Use the uploaded adult man's image as the only face reference and keep his identity precisely intact, including natural skin and expression. Dress him in an oversized maroon shirt worn open over a white vest, loose off white pleated trousers and a simple leather watch, hair slightly windblown and unstyled. Place him on a quiet residential street at night beneath a single warm sodium streetlight, parked scooters and a shuttered shop soft behind him. Have him stand half turned with hands in his pockets, head tilted slightly down and eyes lifted toward the lens, an easy unguarded expression. Light him almost entirely with the warm overhead sodium lamp, letting the light fall off sharply toward the edges of frame. Photograph as if on 35mm colour film, 50mm, f/2, with visible grain, mild halation around the streetlight and slightly soft corners. Grade with warm amber, milky blacks and faded contrast, adding one faint orange light leak in a corner. Vertical 3:4. No text, watermark or digital sharpness.
The intentional imperfection trend is one of the defining shifts of 2026, and it exists as a direct reaction to years of synthetic looking AI images. Grain, halation and milky blacks are the specific vocabulary that produces it. First thing to change: ask for the halation and the light leak to be removed and the same scene becomes a clean modern portrait.
Treat the uploaded photograph of an adult man as the exclusive identity source, preserving facial structure, complexion, age, stubble and expression exactly. Dress him in a charcoal half sleeve knit polo, tailored stone coloured trousers and a slim leather strap watch, everything well fitted and unbranded. Place him on an open apartment rooftop in the last half hour of daylight, a hazy city skyline stretching behind him with a few water tanks and antennae in silhouette. Have him lean back against a low parapet wall with arms lightly folded, one ankle crossed over the other, looking at the lens with a small confident smile. Light him with warm sun low behind camera left creating a defined rim along his arm and hair, with the sky providing soft fill on the shadow side. Photograph as if on a Canon EOS R5 with an 85mm lens at f/2, sharp on the face with the skyline compressed and softened. Grade with warm gold, dusty blue and clean neutral skin, natural contrast. Vertical 4:5. No text, watermark or exaggerated muscle definition.
Well fitted and unbranded does more for a portrait than any amount of styling detail. The closing exclusion matters too, since models left unchecked tend to quietly idealise the body. First thing to change: replace the skyline with hills at the same hour for a travel version of the same shot.
Use the uploaded adult subject's photograph as the sole facial reference. Preserve their exact features, complexion, age, hairline and natural expression, keeping visible skin texture and avoiding any beautification. Dress them in a well fitted navy blazer over a plain white shirt with no tie and no visible branding. Place them against a plain mid grey seamless studio background with no props. Position them squared to the camera from the chest up, shoulders slightly angled, chin marginally forward and down, with a natural closed mouth smile that reaches the eyes. Light them with a large softbox as key at forty five degrees camera left and slightly above, a white bounce card filling the shadow side, and a subtle hair light separating them from the background. Photograph as if on a Canon EOS R5 with a 105mm lens at f/4, both eyes sharp, mild background separation. Grade neutrally with accurate skin tone, no colour cast and no heavy contrast. Vertical 4:5. No text, watermark, glasses glare or over sharpening.
Corporate headshots are the least glamorous item on this list and the most used. The chin forward and down instruction is a working portrait photographer's trick that removes a soft jawline without touching the face. First thing to change: swapping mid grey for a soft white background produces the version most recruiting platforms prefer.
Use the uploaded adult subject's photo as the likeness reference, keeping their recognisable facial features, hairstyle and clothing style identifiable in stylised form. Transform them into a 3D collectible figure with a slightly oversized head, expressive eyes, smooth glossy plastic surfaces and visible moulded seams on the arms. Stand the figure on a round display base beside a glossy retail box with a clear plastic window, the box printed in a bold two colour design with a small blank nameplate area. Place the whole arrangement on a clean wooden desk with a softly blurred bookshelf behind. Light it with a bright three point studio setup producing crisp highlights along the plastic and a soft shadow under the base. Photograph as if on a 100mm macro lens at f/5.6 with the figure and box both sharp. Grade with saturated toy colours, clean whites and a product photography finish. Square 1:1. No readable brand names, no logos and no text on the packaging.
Toyification has been one of the most shared AI photo categories of 2026, driven first by short video and now sitting more permanently on Pinterest. Requesting moulded seams is the small detail that pushes it from cartoon toward convincing plastic. First thing to change: asking for a soft pastel chibi treatment instead of glossy PVC produces the gentler version of the same trend.
Use the uploaded photograph of the adult subject as the only facial reference, preserving exact features, skin tone, age and expression with no smoothing. Dress them in festive traditional clothing in deep maroon and gold with restrained detailing appropriate to the subject. Place them on a home terrace at night during a festival, strings of small warm lights out of focus behind and a low wall lined with earthen diyas in front. Have them cupping a single lit diya at chest height with both hands, face tilted slightly down toward the flame, eyes lowered, a calm and warm expression. Light the face almost entirely from that single flame below, producing a soft warm glow on the cheeks and chin with the background falling into darkness. Photograph as if on a Sony A7 IV with an 85mm lens at f/1.4, the flame sharp and the distant lights rendered as round golden bokeh. Grade with warm amber, deep brown and near black, preserving natural skin colour despite the strong warm cast. Vertical 4:5. No text, watermark or unrealistic flame.
Lighting a face from a single low source is genuinely difficult with a camera and trivially easy to request in a prompt. The instruction to preserve natural skin colour against a strong warm cast is what stops the whole image turning orange. First thing to change: replace the diya with sparklers held at arm's length for a brighter and more energetic frame.
Use the uploaded adult subject's photo as the definitive identity reference and keep the face completely unchanged, including texture and expression. Leave the clothing and setting from the original image broadly intact, adjusting only the photographic character. Re render the image as though it were shot on expired 35mm colour film in the late 1970s and scanned from a physical print. Add fine natural grain, slightly soft focus toward the corners, mild colour shift toward warm magenta in the highlights, and milky rather than deep blacks. Include one faint orange light leak entering from the upper right corner and a barely visible dust speck or two. Keep the subject's face clearly readable and correctly exposed despite the aged treatment. Frame as a standard 3:2 print with very slightly uneven edges. No text, watermark, date stamp or heavy vignette.
This is the least dramatic prompt on the list and quietly one of the most useful, because it converts an existing good photograph rather than inventing a new scene. Explicitly protecting the face from the aging treatment is what separates a stylised result from a ruined one. First thing to change: ask for cooler cyan shifted highlights instead of magenta for a very different decade of film.
| Symptom | Underlying cause | The fix |
| Face resembles a relative rather than the subject | Identity layer missing or too brief | Name the uploaded image as the only reference and list the specific features to preserve |
| Skin looks like plastic | No instruction against beautification | Add a direct instruction to keep pores, fine lines and natural asymmetry |
| Hands look wrong | Hands given nothing to do | Assign every hand a task: resting on a railing, holding a cup, in a pocket |
| Lighting looks flat | Only a mood word was supplied | State direction, source size and colour rather than asking for cinematic light |
| Wrong aspect ratio | Ratio never specified, or inherited from an upload | State the ratio explicitly; when several images are uploaded, the model tends to adopt the ratio of the last one |
| Garbled text in frame | Text left to the model's discretion | Either request no text at all, or specify the exact words wanted |
| Output is a written description, not a picture | Request read as a question | Open with a clear instruction to create or edit an image |
One further habit is worth building. Complex scenes come out better in stages. Generate the base image first, then refine it conversationally with short follow up instructions, which is the multi turn approach Google itself recommends for complicated compositions.
Fourteen prompts covers a month of posting at most. The people who keep this going treat prompts as a collection rather than as one off writing, storing what works, tagging it by occasion, and pulling from that store when a festival or a trend arrives. That habit matters more than any individual prompt.
For anyone who would rather start from a working example than a blank box, PromptSeen maintains a free gallery of copy ready prompts organised by style, with the sample image displayed beside the exact text that produced it. Seeing the output next to its prompt is the fastest way to learn which phrases do the heavy lifting. The trending Gemini photo editing prompts collection is a reasonable place to begin, and the festival and traditional wear sets tend to update as occasions approach.
The deeper benefit of a library is comparative. Reading twenty prompts that all work reveals the shared skeleton beneath them, and after that the skeleton can be filled with anything.
Three points deserve stating plainly, because prompt articles almost never state them.
Every prompt in this article assumes an adult subject working from a photograph they own or have clear permission to use. Generating stylised portraits of other people, particularly anyone who has not agreed to it, is a different activity with different consequences, and no prompt quality makes it acceptable.
Every image Gemini produces carries an invisible SynthID watermark designed to survive cropping, compression and filtering, and Google is expanding support for C2PA Content Credentials, the industry metadata standard for AI provenance backed by Adobe, Microsoft and others. Checking whether an image was AI generated is straightforward: upload it to the Gemini app and ask directly, and the app cross references the SynthID signal. Assuming an edit is undetectable is a mistake.
Disclosure is also becoming a norm rather than a nicety. Several platforms now label AI generated media automatically, and audiences have grown noticeably better at spotting synthetic images. A short note in a caption costs nothing and tends to be received well, particularly for the heavily stylised looks.
Every prompt above will be dated within a year. Feeds move, aesthetics rotate, and the palace courtyard look will eventually feel as of its moment as heavy HDR does now.
The seven layer structure will not date, because it is simply how a photograph gets described: who, wearing what, standing where, doing what, lit how, shot on what, graded which way, minus whatever is unwanted. Learn that sequence and any trend arriving next season becomes a matter of filling in different nouns.
The prompts are the easy part. Noticing what a good photograph is actually made of is the skill worth keeping.
Discussion
Join the discussion and share your thoughts below.
No comments yet. Be the first to share your thoughts!