Banana Prompt XYZ to Copy, Paste, and Post

The phrase banana prompt usually shows up when someone has just watched a reel where a plain selfie turned into a glossy 3D figurine or a moody film portrait, and they want the exact wording behind it. The trailing 'xyz' is junk that search and autofill bolt onto trending terms. It means nothing about the prompt. What you are really after is the set of instructions people feed into Google's Gemini image model, the one the internet started calling Nano Banana. The grid above is stocked with those instructions, written out and ready to lift.

Why your banana prompt search had random letters on the end

When a look blows up, the search box gets messy fast. People half-remember a creator's handle, mash in a tag they saw in a caption, or type whatever autocomplete suggested. So 'banana prompt' picks up suffixes like xyz, pro, v2, new, or a name they can't quite recall. I'd ignore every one of them. The suffix never changes the actual text you paste. Chase the style you saw, not the label. A figurine prompt works the same whether someone filed it under xyz or under their own brand name.

This page is a prompt library, not a person and not an app. There is no secret 'xyz' tool. If you came looking for one specific creator's edit, the honest answer is that the recipe is almost always one of a handful of repeatable styles, and you can run any of them yourself in a couple of minutes.

What a banana prompt actually runs on, in plain terms

Nano Banana is the nickname for Gemini 2.5 Flash Image, Google's image model. You will also see people mention Nano Banana Pro for the sharper, higher-detail jobs. The model takes two things at once: your uploaded photo and your written instruction. It reads both and produces a new image that blends them. That dual input is the whole trick. A normal text-to-image tool only has your words. Nano Banana also has your face, your pose, your lighting, which is why these edits keep the person recognizable instead of inventing a stranger.

It is free to try in the Gemini app on phone and web. Through the API it runs a few cents per image, which only matters if you are batching hundreds at a time. For one-off edits you will never touch a paywall.

Banana prompt looks people are copying right now

Most viral banana edits fall into a small number of buckets. Once you know them, you stop hunting for magic words and just pick a bucket.

Figurine and toy edits

Turn a person into a collectible figure on a desk, sometimes still in the blister pack. These read as fun and giftable, and they forgive a slightly rough source photo because plastic does not need perfect skin.

Retro and film portraits

Warm tones, soft grain, that late-90s point-and-shoot feel. This look leans on lighting words more than anything. Say the era and the film vibe and let the model handle the rest.

Cinematic and editorial shots

Dramatic light, shallow depth, a magazine-cover mood. Couple versions of this are huge for relationship photos. The catch is you have to tell it to keep both faces, or one person quietly gets replaced.

Clean studio headshots

The least flashy and the most useful. Soft even light, plain background, sharp eyes. Good for a profile picture or a CV photo when you don't have time for a real shoot.

How to find the exact banana prompt behind a reel

You rarely need the creator's literal words. You need their recipe. Watch the clip and break the result into four questions, then answer each one in your own prompt.

1. What is the subject? One person, a couple, a pet, an object.
2. What should stay from the original photo? Almost always the face and skin tone.
3. What is the style? Name it: 3D figurine, film grain, studio light, anime.
4. Where does it sit? The background and the lighting around it.

Answer those four and you have rebuilt the prompt without ever seeing it. That is genuinely how most of these spread. Someone reverse-engineers a look, posts their version, and it loops again under a new tag.

Writing or fixing your own banana prompt

A working banana prompt is short. One or two sentences. You write what you want, then add a keep-line so the model does not redraw your face. Here is a quick map from the look you want to the words that get you there.

Notice that every row has a keep-line. Drop it and the model treats your photo as a loose suggestion instead of a base to protect. That single habit fixes more bad results than any clever phrasing.

Tip: a sharp, well-lit input photo beats a clever prompt every time. Sort out the source image first and the edit almost writes itself.

Banana prompt tools: Nano Banana versus the rest

People ask whether they should use Gemini, ChatGPT, Midjourney, or DALL-E for these. Short version: for editing a real photo of a real person, Nano Banana is the easy pick because it actually keeps your likeness. ChatGPT's image tool can do similar edits and is fine if you already live there. Midjourney makes gorgeous images but is built for generating from scratch, so it is weaker when the whole point is preserving your face. DALL-E sits somewhere in the middle. If the task is 'make this photo of me into X', start with Gemini and only branch out if you want a look it struggles with.

Banana prompt mistakes and quick fixes

Most failed edits come from the same handful of slips. Run down this list before you blame the prompt.

• The face changed: your style words were too strong. Soften them and add 'keep the same face and skin tone'.
• Skin looks like wax: add 'natural skin texture' and start from a higher-quality photo.
• Background is a mess: say 'simple background' or name the exact backdrop you want.
• One person vanished in a couple shot: spell out 'two people, keep both faces' near the front of the prompt.
• Every result looks identical and dull: nudge one detail, like the light direction or the era, instead of rewriting the whole thing.

And generate more than one. Results shift slightly each run, so the second or third attempt is often the keeper. Treat the first output as a draft, not a verdict. If two looks are close, run them side by side and let the obvious winner pick itself rather than overthinking the wording.

So forget the xyz. Pick a style from the grid above, upload a clear photo, paste the prompt, and add a keep-line. Run it three times, save the wording that landed best, and you have your own little stash for the next time a new look starts trending.