Google Gemini Retro AI Portraits: Turn Selfies into 4K Bollywood-Style Classics

The retro AI portrait craze, explained

Open Instagram right now and you’ll probably see a familiar face—yours, your friends’, influencers—reimagined as a dreamy 90s poster. Soft grain. Golden-hour light. Chiffon floating like it’s caught in a studio fan. That look is coming from Google’s image features inside Gemini, which creators have nicknamed “Banana” while the underlying mode most people are using is tied to Gemini 2.5 Flash’s image capabilities.

What makes this trend different is how accessible it is. You upload a selfie, describe the style you want, and the model rebuilds the scene with retro textures, film-like color, and stylized wardrobe. People started with 3D figurine edits, then pushed into saree-centric Bollywood looks, old-school mall studio portraits, and poster art that feels like it was shot on Kodak in a small-town photo lab. You don’t need pro gear or editing skills. You just need a clear base photo and a prompt that reads like a director’s note.

The nickname “Banana” stuck because it sounds playful and immediate—exactly how the edits feel. Under the hood, the system blends your reference image with textual directions: lighting, fabric type, camera angle, film stock, grain, and color grading. Multi-turn chats help you iterate: “make the shadows deeper,” “add soft rim light,” “change to chiffon,” “reduce the grain.” The engine responds by re-rendering the portrait without you touching a timeline or a layers panel.

Why sarees? Because the fabric’s movement and texture sell the shot. Chiffon and georgette catch light beautifully, so when you ask for warm directional lighting and a touch of haze, the result feels like a studio setup. Add a textured wall, a vintage lens vignette, and a calm expression, and you’ve got a poster you can almost date to a specific decade. That combination—nostalgia plus control—is why the format rocketed across Reels and Stories.

There are some guardrails. The system aims to avoid harmful or sensitive content, and it can refuse certain celebrity likenesses or unsafe scenarios. But for personal portraits with classic styling, it’s quick, reliable, and remarkably consistent. And because it’s tied to an account you likely already have, the barrier to entry is low. That’s the real story: an edit that used to take a photographer, a stylist, and a retoucher now happens in a chat box.

If you’re wondering about quality, creators are sharing results at high resolution. You can request larger outputs or use built-in super-resolution options where available. When the base image is crisp and well-lit, the final retro portrait often passes as a studio commission—especially once you add period-accurate framing like faded borders or film-burn accents.

How to create your own—and make it look like a real studio shoot

Here’s a simple workflow that mirrors what top creators are doing. It’s fast, repeatable, and doesn’t require design software. You’ll get the best results if you treat it like a mini photoshoot and a script.

First, nail the base image. You don’t need a DSLR, but you do need clarity.

1. Light: Face a window or use a soft lamp at 45 degrees to your face. Avoid overhead lights and harsh shadows under the eyes.
2. Backdrop: A plain wall works. Mid-tone colors (deep green, tan, charcoal) grade nicely into retro palettes.
3. Framing: Shoot from the chest up. Keep the head room tight so the AI has fewer decisions to make.
4. Wardrobe: Solid colors or minimal patterns. You can ask the model to “dress” you in chiffon or georgette later, but clear necklines and shoulders help.
5. Expression: Calm, neutral, or a hint of a smile. Vintage looks rarely go full grin.

Then, run the edit using the image tools inside Google Gemini. The flow many users follow looks like this:

1. Start a new chat and upload your photo.
2. Describe the look in one or two sentences. Focus on fabric, lighting, backdrop, mood, decade, and camera feel.
3. Ask for 2–4 variations to compare (“generate three options with subtle changes in lighting and saree drape”).
4. Refine with short, precise follow-ups: “deepen shadows,” “increase film grain slightly,” “switch to red chiffon,” “add soft rim light on hair,” “keep skin texture natural.”
5. Request a high-res render or ask to upscale the selected version.

Prompts work best when they read like a stylist plus a cinematographer. Think textures, light direction, camera distance, and color temperature. Try these for starters and tweak them to fit your face, hair, and vibe:

• “90s studio poster look, flowing black chiffon saree, deep charcoal wall, warm key light from camera left, soft hair rim light, medium grain, calm expression, subtle vignette.”
• “Vintage mall studio portrait, satin saree with gentle drape, pastel gradient backdrop, soft haze, butterfly lighting, slight film bloom, 4:5 portrait crop.”
• “Romantic red chiffon, sunset-toned background, light wind effect on fabric, classic lens vignette, gentle split lighting, natural skin texture, minimal retouch.”
• “White translucent polka-dot saree, pink flower behind ear, side light with long shadow, muted pastel background, low-contrast grade, diva mood.”
• “Old film still, 1960s palette, sepia-leaning grade, grainy texture, spotlight oval on background, half-profile pose, soft expression.”
• “Retro magazine cover, georgette saree with drape movement, studio fan look, teal-gray backdrop, cinematic bloom, 3:2 aspect ratio.”

Small lines make a big difference. Add camera notes like “shoulder-length crop,” “lens compression feel,” or “headroom tight.” If the model leans too glossy, say “keep pores and fine lines,” “no plastic skin,” or “retain natural lip texture.” To avoid unwanted choices, use negatives: “no jewelry,” “no extra fingers,” “no text in image,” “avoid over-sharpening.”

If you want the exact feel of a 90s photo studio, describe props and placement: “painted canvas backdrop,” “vinyl roll background with subtle wrinkles,” “hand resting on a pedestal,” “faint paper texture on edges,” “glossy print look.” These details lock in realism because they mirror what real studios used back then.

Common hiccups and how to fix them:

• Fabric looks stiff: Ask for “chiffon” or “georgette,” and add “light movement, delicate folds.”
• Skin looks airbrushed: Say “keep natural skin texture, minimal retouch, visible fine details.”
• Colors are too modern: Ask for “muted film palette, warm highlights, cool shadows, gentle contrast.”
• Background feels flat: Add “textured wall,” “spotlight oval,” or “gradient studio paper.”
• Face shape shifts: Re-upload a sharper selfie and specify “preserve likeness, subtle stylization only.”

Aspect ratio and framing matter if you’re posting. For Instagram feed, 4:5 (1080×1350) fills the screen. For Stories and Reels, 9:16 works best. If you plan to print, try 3:2 or 4:3 and ask for “clean edges, no heavy vignette.” When you request a larger export, mention “high-resolution portrait” or “upscale to near-4K” so the model prioritizes detail in hair strands, eyelashes, and fabric weave.

The signature lighting setups for this trend are simple. Three to try:

• Warm key from the left, soft rim from the right, faint hair light from behind. This creates depth without crushing shadows.
• Butterfly light (soft light above the lens) for a classic beauty look with a gentle shadow under the nose.
• Split light (key at 90 degrees) for drama. Add a bounce fill to keep it printable.

Not into sarees? The same approach works for retro tuxedos, leather jackets with 80s neon, or 70s studio fashion with flared pants. Use wardrobe descriptors like “satin lapels,” “wool blend,” “denim fade,” or “silk scarf.” For film references, try “Agfa-leaning red,” “Kodachrome-inspired saturation,” or “bleach-bypass style contrast.” The model responds well to classic film language.

Creators are also pairing two images—their selfie plus a reference frame—to steer composition. If the tool allows an additional reference, upload a vintage poster and say, “match lighting, crop, and color mood, adapt wardrobe to saree style.” This nudges the system toward a faithful homage without copying.

Worried about privacy and ethics? Stick to your own photos or get clear consent if you’re editing someone else’s image. Label AI art in your caption if your audience expects transparency. Avoid misleading edits of public figures. And be thoughtful with cultural styling—respectful references land better than mashups that ignore context.

How does this compare with other AI art tools? Dedicated image models like Midjourney and Stable Diffusion excel at stylization and cinematic lighting, but they often require Discord prompts or local setups and upscalers. Mobile apps are fast but can feel template-driven. Gemini’s advantage right now is convenience: conversational edits, solid likeness preservation when the input is clean, and quick refinement inside a single chat. If you want deep control over nodes, masks, or LoRAs, other ecosystems still have more levers. If you want a gorgeous portrait fast, this is hard to beat.

To push quality even further, add a finishing pass. After you pick your favorite render, ask for “micro-contrast on fabric only,” “soft highlight recovery on forehead,” or “reduce color noise in shadows.” If details get crunchy, say “gentle denoise, keep grain.” And if you plan to print, request “true blacks not crushed, print-safe contrast.”

Here are a few targeted prompt modules you can copy-paste into your request when you need a specific tweak:

• Lighting precision: “key light 45° camera left, 3200K warm; fill at 20%; hair light subtle; background kicker dim.”
• Fabric realism: “fine chiffon translucency, soft specular highlights, layered drape with airy motion.”
• Color grade: “muted film palette, warm mids, slight teal in shadows, gentle halation on highlights.”
• Lens vibes: “vintage 85mm look, light vignette, mild chromatic aberration at edges.”
• Detail control: “natural skin pores, clean catchlights, no plastic smoothing, retain fine hair.”
• Clean output: “no text in image, no watermark, no logos, no extra fingers, hands cropped if unsure.”

For the retro mall studio aesthetic, call out the telltales: “airbrushed gradient backdrop, soft front fill, glossy print feel, slight paper curl at corners, 1990s portrait studio style.” Want that family-album charm? Ask for “matte photo border, date stamp in corner, faded ink look,” then remove the date if it looks too on-the-nose.

If your results feel too modern, push further into period detail: “bronze makeup tones, thin eyeliner, softly feathered brows, muted lipstick, 90s hair volume.” For men’s portraits, try “gelled side part, soft fall-off light, velvet blazer, subtle grain, warm sepia lean.” Small cosmetic cues anchor the decade.

Resolution talk, without the jargon: you can get crisp results if your selfie is sharp. Avoid heavy compression. If you see smudging in hair or fabric, ask for “re-render at higher detail” or “upscale with improved edges.” And when exporting for Instagram, keep the long edge near platform defaults so the app doesn’t crush it with aggressive compression.

The community side of this trend is moving fast. People are sharing prompt cards, style recipes, and before/after carousels. Photographers are mixing AI with real lights: shoot a decent key light, then let the model rebuild the rest. Some creators are selling preset packs—prompt bundles with specific lighting, color, and fabric notes—so clients can request a “Black Chiffon 90s,” a “Polka-Dot Diva,” or a “Sunset Red Romance” edit on demand.

Expect the look to expand beyond sarees. We’re already seeing retro campus portraits, old film set posters, and glam studio shots with glossy backdrops and dramatic shoulder pads. The tools are getting better at hands, jewelry, and hair flyaways; the prompts are getting more precise; and the audience still loves a good throwback when it looks like it came from a real photo album.

If you only remember one thing, it’s this: your words are the camera. Be specific about light, fabric, mood, and frame. Give the model a clear, well-lit selfie. Iterate in small steps. And when you hit the sweet spot—a calm, confident expression, warm directional light, and fabric that breathes—you’ll understand why these portraits are taking over feeds right now.