90s AI Photo Trend: Exact Prompts for Gemini and ChatGPT

By Fernando Lisboa | | Updated , 6 min read

90s AI Photo Trend

The 90s AI photo trend is spreading fast, and the two tools most people use for it behave very differently. One photo, one detailed prompt, and you get a convincing throwback snapshot in seconds, but only if the prompt matches the tool.

Gemini handles real faces combined with vintage scene details without tripping safety filters as often, so it is the easiest starting point. ChatGPT is stricter around real human faces, and the same prompt that works on Gemini often gets watered down or rejected there.

Below are both prompts, structured the way a prompt engineer would build them, plus the small adjustments that separate a convincing photo from an obviously fake one.

The Exact 90s AI Photo Prompt for Gemini

Attach a clear, front-facing photo of yourself, then paste this:

Prompt
Recreate this photo as a real 1990s point-and-shoot snapshot (not 1980s). Subject: an adult man with brown/tan skin tone, natural candid expression, sitting on a couch. Action: playing a classic gray home video game console, holding a boxy controller with both hands, eyes focused on the TV. Setting: 1990s living room, wood paneled walls, a warm yellow lamp, a bowl of popcorn and a mug nearby, generic vintage video game posters on the wall. Wardrobe: bold patterned knit sweater typical of the early-to-mid 1990s. Camera and film quality: visible film grain, warm amber-orange color cast, harsh direct on-camera flash, natural skin texture, slightly soft focus, minor motion blur, small timestamp in the corner. Avoid: modern consoles, modern furniture, digital sharpness, studio lighting, airbrushed skin, visible logos or brand names, anything from the 1980s specifically.

✅ Breaking the prompt into Subject, Action, Setting, and Wardrobe blocks keeps Gemini from skipping details buried in a long paragraph
✅ “Not 1980s” locks the decade, since Gemini sometimes blends neighboring eras when you only say “90s”
❌ Skip brand names entirely, generic descriptions like “gray home video game console” work better than naming one

Edit the Subject line to match yourself. Gemini follows the photo for the face, so the description only needs to cover what the photo does not show.

The Exact 90s AI Photo Prompt for ChatGPT

ChatGPT’s image model runs a stricter check on prompts that combine a real uploaded face with detailed physical descriptions of that same face. The workaround is to shift the weight of the prompt: describe the room, not the person. ChatGPT then spends its attention on the video store, the lighting and the grain, and treats the face as a simple carry-over from the photo.

Attach a clear, front-facing photo of yourself, then paste this:

Prompt
Turn this photo into a candid snapshot taken inside a 1990s American video rental store. Subject: keep the person’s face and features from the attached photo, natural unposed expression, standing in one of the aisles. Setting: rows of VHS tapes on shelving, hand-written cardboard genre signs, checkered tile floor, fluorescent overhead lighting, a return bin near the counter. Wardrobe: casual early-to-mid 1990s clothing, an oversized graphic tee or a windbreaker jacket. Camera and film quality: visible film grain, slightly warm color cast, on-camera flash with mild red-eye, soft focus, small corner timestamp reading 1994. Avoid: modern signage, digital displays, smartphones, airbrushed skin, studio lighting, visible store brand names or logos.

✅ Naming the location first gives ChatGPT a scene to build around instead of a face to scrutinize, which lowers the chance of a refusal
✅ “Keep the person’s face and features from the attached photo” works better than re-describing the face, since ChatGPT already has the reference image
❌ Never write “photorealistic” in the prompt. ChatGPT associates that word with deepfake requests and blocks more often when it appears

If the first result looks off, reword the scene before touching the wardrobe or lighting. Small changes, like swapping “video rental store” for “video store aisle,” sometimes change the outcome more than any lighting detail. The same scene-first idea powers the 3D figurine AI trend, which hits the same refusal wall.

Tips to Make the Result Look Authentic

  • Ask for “slight underexposure” instead of even lighting, real flash photos from that era were rarely well balanced
  • Request a “cropped, off-center composition,” since candid 90s shots were almost never centered
  • Add “one slightly blurry hand” or “flash reflection on a nearby glossy surface,” details almost every disposable camera photo from that decade has
  • Never add “smooth skin” or “perfect lighting” to the prompt, that instantly makes the result look digital

Run the prompt two or three times if the first output feels too polished. Even with identical wording, both tools vary their output, and a second attempt often nails the grain and flash texture better than the first. Want more prompt formats like these? Our guide to 100 ChatGPT image prompts covers lighting, framing and camera angles, and our Grok Imagine 2.0 prompt formula works well for text-heavy scenes. For a wider comparison of tools, see the best AI image generators of 2026.

ChatGPT vs. Gemini for the 90s AI Photo Trend

Feature🤖 ChatGPT🎨 Gemini
👤 Real face handling⚠️ Stricter, needs indirect wording✅ More permissive
⏱️ Average attempts needed2 to 31 to 2
🎞️ Film grain accuracy✅ Strong✅ Strong
💰 Free tier available✅ Yes✅ Yes
🚫 Refusal rate on first tryHigherLower
🖼️ Best use caseScene-heavy shots (stores, arcades)Close-up portrait shots

FAQ

Does Gemini require a paid subscription for this?

No. Gemini offers a free tier that handles this type of prompt well. Results are consistent enough for casual use, though a paid plan may offer faster processing during high traffic periods.

Does ChatGPT refuse this prompt every time?

No. Refusals happen more often than on Gemini, but leading with the setting instead of the face lowers the refusal rate significantly. If it fails once, rephrase the location detail and try again.

Can I use the Gemini prompt in ChatGPT word for word?

Not reliably. The Gemini prompt describes the person in detail, which ChatGPT reads as a stricter request. Use the scene-first version above instead.

Why did my photo come out too clean?

Overly sharp or well-lit source photos confuse the AI’s sense of “vintage.” Use a photo with natural, slightly uneven lighting, and add “slightly underexposed” and “visible flash glare” to the prompt.

Which tool gives better results for this trend overall?

Gemini currently produces convincing results faster and with fewer refusals. ChatGPT catches up when the prompt centers on a detailed scene rather than a close-up face.

Try It Before the Trend Fades

Grab a recent selfie, copy the prompt for your tool into Gemini or ChatGPT, and post your result while this wave is still peaking.

Sources