A useful Gemini AI photo prompt tells the model what to make, what to keep and what to change. You can copy the prompts below, replace the bracketed details and use them for a new image or an edit of your own photo.
For a familiar image workspace, open Nano Banana 2 in ShimmerPic. For Google's own interface, start at Gemini. The instructions below explain both paths.
How to use Gemini AI photo prompts
In Gemini, choose image creation, enter your prompt and generate. To edit a photo, upload it and describe the changes. Google's image-generation help explains the current controls and account availability.
In ShimmerPic, open the Nano Banana model page, paste the prompt and add a reference image when the example needs one. Choose an available aspect ratio and resolution, check the displayed credit cost and generate. Keep the first instruction focused; you can refine the result afterwards.
For a reference-led edit, explicitly name the parts that should stay recognizable. A request to keep an object unchanged is a useful instruction, but you should still compare the result with the original before using it.
Gemini AI photo prompts to copy and paste
These examples cover different tasks. The portrait, product, figurine, knitted-doll and room prompts need a reference photo. The poster and landscape prompts can start from text alone.
1. A natural editorial portrait
Use a clear photo of yourself or someone who has agreed to the edit. Pick a scene that changes the setting without demanding an entirely new face.
Edit the uploaded portrait into a natural editorial photograph. Keep the person's facial features, hairstyle and approximate age recognizable. Place them beside a large window in a quiet [studio or cafe], wearing a [cream linen shirt]. Use soft daylight, gentle shadows and realistic skin texture. Frame from the chest up. Keep the expression relaxed. Avoid heavy beauty retouching and add no text.
If the result looks overly polished, ask for less skin smoothing and more natural lighting before changing the setting again.
2. A product photo with a new background
Start with a sharp product image on a simple background. Mention the details that matter to the listing.
Use the uploaded [product] as the subject of a clean product photograph. Preserve its shape, proportions, material, label wording and visible details. Place it on a pale stone surface with [soft green leaves] in the background. Light it from the left with a large soft source. Keep the entire product visible and leave empty space on the right for a caption. Do not add accessories attached to the product or invent new branding.
Inspect labels, seams and packaging closely. For more scene ideas, explore our AI product photography page.
3. A miniature figurine scene
This describes an illustration of a figurine, not a manufacturing-ready model or a physical product.
Turn the subject in the uploaded photo into a small collectible figurine on a plain circular base. Keep the outfit colors and hairstyle recognizable. Put the figurine on a tidy wooden desk beside an unbranded sketchbook. Use softly painted resin textures, a slight overhead camera angle and warm window light. Show the complete figurine and its base. No logos, packaging claims or text.
For a cleaner result, remove extra desk objects from the prompt rather than adding more detail to every surface.
4. A hand-knitted character
Reimagine the subject in the uploaded photo as a small hand-knitted doll. Preserve the recognizable hairstyle, glasses if present and main clothing colors. Use visible wool loops, simple embroidered features and rounded proportions. Place the doll upright on a light wooden shelf, photographed in soft daylight. Keep the full doll inside the frame, with no text or extra characters.
If the model adds accessories, make the next instruction specific: remove the hat while preserving the doll, pose and background.
5. A poster with a short headline
Create a portrait-format poster for a fictional neighborhood ceramics workshop. Use warm ivory paper, cobalt blue ceramic shapes and a restrained orange accent. Place the exact headline "MAKE A LITTLE SPACE" near the top, large and easy to read. Below it, include only "Saturday pottery club". Leave generous margins and keep the text separate from the objects. Add no other wording.
Short copy is easier to inspect than a page full of small print. Check every letter. For a text-focused starting point, try Nano Banana Pro and review the lettering before publishing.
6. A room makeover that keeps the layout
Edit the uploaded room photo into a calm reading corner. Keep the camera angle, window positions, doors and room dimensions visually consistent. Change the sofa fabric to [muted olive], add one warm reading lamp and use a natural woven rug. Preserve the existing furniture placement. Use believable evening lighting. Do not add windows or remove walls.
Treat the image as a visual concept. Furniture fit and building measurements still need to be checked separately.
7. A cinematic landscape from text
Create a cinematic photograph of a small coastal train station just after rain. Pale blue benches sit beside the platform, with warm light from a single ticket-window lamp. Show wet stone reflections, distant sea mist and a quiet early-morning atmosphere. Use a wide composition with the station on the left. No people, signs, logos or text.
Fix an image with one clear follow-up
When a result is close, keep the successful parts in your next instruction:
Keep the subject, composition and lighting from this image. Change only [the background to a plain warm-gray wall]. Preserve [the face, product label and outfit]. Do not add new objects.
Change one thing at a time. If both the face and the entire scene are wrong, return to the reference and simplify the original request instead of stacking corrections on top of it.
Which Nano Banana model should you choose?
ShimmerPic offers Nano Banana, Nano Banana 2 and Nano Banana Pro on their own model pages. Start with the model and settings that fit your task, then compare a more detailed option if the layout or lettering needs work. Available settings and credit costs are shown before generation.
The model directory is the place to compare options. The prompt library adds community examples when you need a different visual direction.
Common questions
Can I use these prompts in Gemini and ShimmerPic?
Yes. The descriptive wording can be used in either interface. Choose a suitable image model and upload any reference the prompt calls for. Controls, availability and results can differ between services.
Why did my face or product change?
Image models can alter details even when asked to preserve them. Use a sharper reference, name the important features and reduce the number of simultaneous changes. Reject a result that changes a detail essential to the image's purpose.
Does a longer prompt always work better?
No. A clear subject, a specific change and a few visual constraints are a better starting point than conflicting style instructions. Add detail only when it resolves something the model got wrong.
Pick one example, replace its bracketed details and make your first version with Nano Banana 2.