Nano Banana 2 retro photo prompts: 6 controlled edits

You want a retro photograph, but adding the word vintage can change more than the look. The face, garment or product may drift while the model adds grain and a color cast.

A useful edit needs a clear visual direction and a list of details to preserve. This guide gives you six controlled prompt routes, from warm studio portraits to direct-flash snapshots and catalog images. You’ll see how to describe the medium, light and era-specific texture, then compare the result with the source before you continue from an approved version.

Retro edit workflow

1. Choose one era and medium. A 1970s studio portrait, 1990s point-and-shoot and Y2K catalog image need different light, color and texture.

2. Prepare the source. Use the highest-quality original you are allowed to process; crop only after deciding the delivery ratio.

3. Paste one prompt. Keep the subject and scene constraints explicit instead of stacking several aesthetics.

4. Inspect the full image. Check identity, hands, hair, clothing seams, text, logos, reflections, objects and background.

5. Correct one problem at a time. Continue from the last approved version and name the exact region to repair.

6. Export and document. Save the original, prompt, model, selected output, edits, rights and date.

A reusable preservation suffix

Append this to any prompt when those properties should remain stable:

“Preserve the subject’s identity, facial features, expression, pose, hands, body proportions and clothing shape. Preserve the camera angle, crop, background layout and all existing readable text. Change only the photographic treatment described above. Add no logos, jewelry or objects.”

A preservation instruction improves reviewability but does not guarantee exact fidelity. Compare every output with the original at full resolution.

Inputs and first-attempt results

We ran each of the six prompts below once on November 3, 2026 in Magic Hour with Nano Banana 2 selected, 1K resolution and the source aspect ratio. The six completed runs used 600 credits in total. The portrait is a fictional adult; the product source is an unbranded coffee arrangement. Results are shown without retries, including a prompt-following failure. These examples are not a model benchmark.

Portrait input for examples 1–4 and 6

1. 1970s warm studio portrait

First-attempt result

The first result adds a warm cast, grain and softer contrast. Skin color and texture also change, so compare the face with the source before accepting this treatment.

Check: skin tone, highlight clipping, garment color, invented accessories and whether the backdrop changed shape.

2. Direct-flash 1990s snapshot

Prompt: “Make this look like a late-1990s compact-camera print scanned from a family album. Use direct on-camera flash, a slightly dark background, modest red-eye risk without distorting the eyes, consumer-film color, fine print grain and a narrow white photo border. Preserve every person, face, hand, pose and object. Add no timestamp unless requested.”

First-attempt result

This first attempt added a date and border text even though the prompt says to add no timestamp. The date is invented. Remove it before using the picture; a preservation instruction is not a guarantee.

Check: eye detail, flash shadows, background people, fingers and any invented date or camera mark.

3. 1980s editorial portrait

Prompt: “Transform only the photographic treatment into an early-1980s editorial portrait. Use a controlled hard key light, subtle rim light, deep but readable shadows, restrained color saturation, crisp medium-format detail and light print texture. Preserve the subject’s identity, hairstyle shape, pose, wardrobe construction and background geometry.”

First-attempt result

The first result makes a subtler change than the prompt suggests. Compare the light, hair edges and fabric texture with the input rather than assuming the requested decade is visible.

Check: face shape, hairline, fabric edges, shadow detail and whether the model introduced era stereotypes you did not request.

4. 1960s black-and-white contact print

Prompt: “Render this as a carefully printed 1960s black-and-white documentary photograph. Use panchromatic monochrome response, moderate contrast, visible but fine silver-halide grain, natural edge softness and a thin contact-print border. Preserve the exact people, scene, signage and crop. Do not simulate damage, scratches or dust.”

First-attempt result

The first result is monochrome with a contact-print border. Check shadow detail and the face against the original; the border itself does not establish photographic authenticity.

Check: readable signage, facial separation from the background, lost shadow detail and fake damage.

5. Y2K catalog product photo

Prompt: “Restyle this product image as a clean early-2000s mail-order catalog photograph. Use an off-white seamless background, broad softbox light, small natural contact shadow, slightly cool color balance and restrained digital sharpening. Preserve the product shape, label, logo, color, material and proportions exactly. Add no copy, badges or packaging.”

Product input for example 5

First-attempt result

The result restyles the coffee arrangement on a light catalog background. Compare bag, cup and bean shapes with the input. This unbranded source does not test preservation of label text or a logo.

Check: label spelling, logo geometry, product dimensions, material texture and new claims or badges.

6. Faded instant-film portrait

Prompt: “Create the look of an aged instant-film print without changing the scene. Use a soft central exposure, mild cyan shadows, gently faded warm highlights, low micro-contrast and a wide blank instant-film border. Preserve the subject’s identity, hands, clothing, pose and background. Do not add handwriting, dates, light leaks or heavy damage.”

First-attempt result

The result adds an instant-film border and a faded treatment. The border is visibly weathered; inspect whether that amount of aging suits the intended use and whether the face remains acceptable.

Check: face fidelity, hand anatomy, clothing color, border shape and invented writing.

How to repair a near-miss

• Identity drift: restart from the source and reduce the requested change to light, color and texture.

• Unwanted props: list the additions to remove and repeat that no objects may be introduced.

• Damaged text or logo: restore it from the original in a deterministic editor or mask it from the generative edit.

• Too much grain or damage: ask for fine grain and a clean professional scan, then prohibit scratches, dust and light leaks.

• Wrong decade: describe the camera, light, palette and print process; the year alone is too ambiguous.

• Repeated-edit degradation: return to the last approved file instead of correcting a correction.

Choosing the right Nano Banana model

Nano Banana 2 is Gemini 3.1 Flash Image, now the previous-generation model in Google’s image family. Google recommends Nano Banana 2.1 (gemini-nano-banana-2.1) for new projects. The original Nano Banana refers to legacy Gemini 2.5 Flash Image. These retro-photo prompt recipes remain starting points to adapt to the exact model exposed by your tool, rather than a measured comparison of model versions. Original guide checked September 13; model selection re-checked October 10, 2026.

Google’s current image-generation guide recommends Nano Banana 2.1 for new general work, Nano Banana 2 Lite when speed and cost dominate, and Nano Banana Pro for complex professional assets. Google labels the original Gemini 2.5 Flash Image model as legacy. The Gemini Apps help page explains current app access, age, plan and regional limits.

For the exact model IDs, app steps and multi-reference workflow, use the complete Nano Banana 2 tutorial.

When another workflow fits better

Use Magic Hour’s AI Image Editor for a guided source-image edit, or the AI Image Generator when you want a new retro scene without a source photo. Browse AI art styles for medium vocabulary. After the still passes review, use Image-to-Video if the final asset needs motion.