Most Gemini photo trends work the same way: you upload your own photo, paste one prompt, and get yourself back as a figurine, a film still, a magazine cover or a festive portrait. When it works, it looks like you. When it does not, you get a stranger with your haircut.

The difference is almost always the prompt. This guide covers how Nano Banana reads prompts, the lines that keep a face consistent, and a set of example prompts you can paste into Gemini or run in the studio.

How Nano Banana reads a prompt

Nano Banana is a language model that draws. It reads your prompt the way it would read a request in a chat, so it responds to full sentences, not keyword lists. A string like "portrait, bokeh, 85mm, cinematic, 8k, masterpiece" gives it fragments to guess at. Two to four plain sentences that describe the subject, the setting, the light and the camera give it a scene to build.

A reliable order:

  1. Subject and action. Who is in the picture and what they are doing.
  2. Setting. Where it happens, with one or two specific details.
  3. Light. Direction and quality: soft window light from the left, low warm sun behind, a single flash.
  4. Camera or lens. Close-up or full body, eye level or low angle, shallow depth of field.
  5. Style. Photo, film still, oil painting, 3D render.

Leave out filler such as 8k, ultra detailed or award-winning. It does not change the result and it crowds out the words that do.

The identity line

When you edit your own photo, the model has to decide what to keep. Without instructions, it may smooth the skin, change the jawline or shift the eye colour while it restyles everything else. One sentence fixes most of this:

Nano Banana
Keep the person's face, facial features, hairstyle and skin tone exactly as in the uploaded photo.

Put it at the start or the end of every edit prompt. For couples and families, say "every person's face" so it applies to everyone in the frame.

Example prompts to copy

Each of these assumes you upload one clear, front-facing photo. Replace anything in square brackets.

Studio headshot from a selfie

Nano Banana
Turn this selfie into a professional headshot of the same person wearing a charcoal blazer over a white shirt, standing in front of a plain light gray studio backdrop. Soft key light from the front left with a gentle fill, shoulders angled slightly, looking into the camera with a relaxed smile. Shot at eye level with an 85mm lens and shallow depth of field. Keep the face, hairstyle and skin tone exactly as in the photo.

The library has a tested version of this in LinkedIn headshot.

Film still at night

Nano Banana
Place the person from the photo on a rain-wet city street at night, walking toward the camera with hands in the pockets of a long dark coat. Neon shop signs in teal and magenta reflect on the pavement behind them. Medium shot at eye level, shallow depth of field, the look of a still from a 1990s crime film with soft grain. Keep their face and hairstyle exactly as in the photo.

Desk figurine

Nano Banana
Create a 1/7 scale collectible figurine of the person in the photo, standing on a clear round acrylic base on a wooden desk. Behind it, a computer monitor shows the 3D model of the same figurine, and a printed box with the figurine's picture sits beside it. Bright, even desk lighting, product photo style. Keep the face recognisable as the person in the photo.

A version of this trend is in the library as 3D collectible figurine of yourself.

Fixing a real photo

Edits are where Nano Banana is most useful day to day. Ask for one change per prompt:

Nano Banana
Relight this photo as if it were taken in the last hour before sunset: warm, low sunlight from the right, long soft shadows and a golden tint on the skin and hair. Do not change the people, their poses, their clothes or the background layout.

More edits like this, from restoring creased prints to removing people from the background, are in the photo editing prompts hub.

Text inside the picture

Posters, cards and covers need words. Put the exact text in double quotes and say where it goes and how it looks:

Nano Banana
A fashion magazine cover with the person from the photo as the cover star, three-quarter portrait against a warm beige backdrop. The masthead "[MAGAZINE NAME]" runs across the top in large white serif letters, and the cover line "[YOUR NAME]: the interview" sits in the lower left in bold sans serif. Glossy print look. Keep the face exactly as in the photo.

Nano Banana Pro is noticeably better at spelling and placing text than the original model. Keep the text short; a headline and one line work better than a paragraph.

When the result is wrong

  • The face drifted. Add or strengthen the identity line, and ask for fewer changes at once. Restyling the outfit, the setting and the light in one go gives the model more room to wander.
  • The colours look oversaturated. Name the palette you want ("muted, natural colours, true-to-life skin tones").
  • It ignored part of the prompt. Move the ignored part to the first sentence. Early details tend to carry more weight.
  • It changed things you wanted kept. List them: "Do not change the background, the clothes or the pose."
  • The aspect ratio is wrong. Say it in words: a vertical 4:5 portrait, a wide 16:9 scene. In the studio you can also set the ratio directly.

Nano Banana, Nano Banana 2 or Nano Banana Pro

All three read the same kind of prompt, so you do not need to rewrite anything to switch. The original Nano Banana is fast and good at everyday edits. Nano Banana Pro is the one to pick for text, crowded scenes and composites from several photos; in the studio it accepts up to six reference images, against three for the original. Nano Banana 2 adds sharper detail than the original and takes up to six reference images too. If a prompt works on one, try it on another before you start rewording.

Write your own in a minute

The image prompt generator writes Nano Banana prompts in this structure from a one-line idea: pick Nano Banana as the target, choose a style and aspect ratio, and you get full sentences with the light and lens already chosen. If you already have a photo whose look you want to copy, image to prompt writes the prompt from the picture.

For ready-made prompts, browse the Gemini photo prompts hub and the couple photo prompts, or the whole image prompt library.