Nano Banana AI Photo Prompts: The Complete Guide

Nano Banana is the name that stuck for Google's image-generation and image-editing model inside Gemini, and by 2026 it has become the tool most people mean when they search for a “trending AI photo prompt.” Its popularity did not come from being the flashiest option on paper. It came from being unusually good at one specific thing people actually want: taking a real photo of a real face and changing the styling, outfit, setting or format around it while keeping that face recognizable.

That distinction matters more than it sounds. Most earlier image generators were built to create a scene from scratch out of a text description. Nano Banana behaves more like an editor you talk to in plain language. You give it a photo, describe the change you want, and it applies that change while treating the rest of the photo as something to preserve rather than something to reinvent. Once people realized this, the entire category of “turn my selfie into X” prompts took off.

This changes how you should think about writing a prompt for it. Instead of describing an entire scene as if nothing exists yet, describe the transformation. State what should change, state what should stay the same, and be explicit about the one thing most people forget: that the person's actual face is the anchor the whole image is built around, not a detail to be regenerated.

A dependable Nano Banana prompt usually has four parts. First, name the transformation directly — a 3D figurine, a traditional saree portrait, a retro film poster, a polaroid with someone else in frame. Second, describe the specific visual details of that transformation: textures, lighting, props, framing. Third, state the identity instruction on its own line so it does not get lost among styling details. Fourth, mention the output format if it matters, such as a square frame, a visible border, or a plain background.

The identity instruction deserves its own sentence rather than being folded into a longer description. Something like “preserve the subject's natural facial structure, skin tone and features exactly as shown in the reference photo” does more work than a dozen aesthetic adjectives. Nano Banana tends to respect an isolated, clearly stated constraint more reliably than one buried in the middle of a stylistic paragraph.

Most of the trend categories currently spreading on Instagram are really the same underlying skill applied to different concepts. The collectible-figurine trend asks Nano Banana to restyle a person as a toy while keeping their face. The traditional-saree trend asks it to change clothing and setting while keeping the face. The retro Bollywood-poster trend adds typography-style framing and film-grain texture on top of the same core instruction. Once you understand the pattern, you can adapt it to a concept nobody has tried yet.

The polaroid-style trends work slightly differently because they often combine two reference photos — a present-day photo and a childhood photo, or photos of two different people. Here it helps to name each subject's role explicitly (“the adult in the first reference,” “the person in the second reference”) so the model does not blend two identities into a single face, which is the most common failure in these prompts.

A workable structure is: name the transformation → describe the specific visual details → state the identity-preservation requirement on its own → specify format or framing if relevant. Keep each part to a sentence or two. Nano Banana responds better to a short, well-ordered prompt than to an exhaustive one.

If the face starts drifting from the reference, the identity sentence is usually too vague or too far from the front of the prompt — move it earlier and make it more specific. If the styling looks weak or half-applied, the transformation description is probably too short; add concrete visual details like texture, lighting direction and color rather than single adjectives. If two people's faces start blending in a multi-reference prompt, name each person's role explicitly instead of referring to “both people” as a group.

One underused feature of working with Nano Banana through Gemini is that you can keep the conversation going after the first result. Rather than rewriting the whole prompt because the lighting is slightly off, describe only the adjustment you want — “make the lighting warmer” or “move the light source to the left.” This mirrors how a real photo edit works and usually gets you to a usable result faster than regenerating from scratch each time.

Once a Nano Banana prompt produces a result you like, treat it as a template rather than a one-off. Swap the reference photo, keep the transformation instruction the same, and you have a repeatable format for a Reel series. This is exactly how most of the current viral trend formats spread — not because one image went viral, but because the underlying prompt structure was easy for anyone to reuse with their own photo.

CreateLoom's Nano Banana-compatible prompts are written with this structure in mind, so each one separates the transformation, the visual details and the identity instruction clearly — copy one, drop in your own reference photo, and adjust only the part that needs to change for your version of the concept.