AI prompt library
Free nano banana prompts ready to paste. Drop one into Gemini or ChatGPT, add your photo, and see what comes back.
15 prompts here. Free, no signup.
A nano banana prompt is the line of text you paste into Google's Gemini image tool to edit a photo or build one from scratch. People nickname the model nano banana ; the real name is Gemini 2.5 Flash Image. The grid at the top of this page is full of prompts you can copy right now. This part is about the words themselves: how a good prompt is built, and how to fix the ones that come out wrong. What a nano banana prompt…
Moon-terrace karwa men, moonlight, channi or thali as the title says, rooftop or courtyard, respectful ritual. Copy paste Nano Banana look for men with face lock.
A clean studio portrait in black tee. Copy paste Nano Banana prompt for men with face lock and seamless grey backdrop.
A heritage corridor portrait in cream shirt. Copy paste Nano Banana prompt for men with face lock and stone arches.
A night market portrait in denim jacket. Copy paste Nano Banana prompt for men with face lock and warm shop glow.
A rooftop portrait at golden hour in white shirt. Copy paste Nano Banana prompt for men with face lock and warm sun.
A wool coat portrait on a grey overcast street. Copy paste Nano Banana prompt for men with face lock and soft daylight.
Create a photoreal office jetpack meme with a fictional tech worker floating above coworkers, believable materials, and cinematic 4:5 framing.
Retro 80s prompt with synthwave energy: glowing arcade cabinets, neon pink and cyan light, a varsity jacket and heavy film grain. Copy, paste, create.
Create a glowing DNA hologram portrait with a fictional scientist, realistic hands, cinematic lab lighting, and polished vertical 4:5 detail.
Create a viral surreal identity portrait with glass fragments, luminous neural details, realistic skin, and cinematic 4:5 framing for Nano Banana.
Create a cinematic first-person shooter concept with neon rain or desert ruins, realistic game lighting, and vertical 4:5 composition.
Create a fictional AI rivalry portrait with three executives, dramatic boardroom lighting, clear power dynamics, and premium editorial realism.
Create a low-angle urban portrait with a sharp seated subject, radial crowd motion blur, golden-hour light, and cinematic orange-and-teal grading.
Create a luxury elevator couple mirror selfie while preserving both uploaded facial identities, natural expressions, skin texture, and authentic details.
Recreate a colorful outdoor fashion portrait with exact face identity, a locked full-body pose, and a matching anime chibi mural.
No prompts match your search.
The guide
A nano banana prompt is the line of text you paste into Google's Gemini image tool to edit a photo or build one from scratch. People nickname the model nano banana; the real name is Gemini 2.5 Flash Image. The grid at the top of this page is full of prompts you can copy right now. This part is about the words themselves: how a good prompt is built, and how to fix the ones that come out wrong.
The model reads your photo and your instruction at once, so a prompt is half description, half rules. You tell it what to make and quietly tell it what to leave alone. That second part is where people slip: they describe a lovely scene but forget to say 'keep my face', so the output looks great and nothing like them. Plain English works best, no brackets or JSON-style formatting needed; a clear sentence beats a wall of fake code.
Good prompts share the same skeleton. Write yours in this order:
Stitch those into one sentence: a portrait of me, keep my exact face, warm 35mm film look with soft grain, by a window in afternoon light, shot close.
Tip: decide the keep line before the look line. Nail the likeness first and you can throw any style at it; chase the style first and the face drifts.
The looks shared most lean editorial and filmic. Pick a row, copy the wording into Gemini, swap in your own photo.
| Style | The vibe | Wording to try |
|---|---|---|
| Analog film portrait | nostalgic, grainy, warm | '35mm film, soft grain, faded warm color, keep my face' |
| Soft editorial portrait | magazine, clean, posed | 'editorial lighting, muted tones, sharp focus on the eyes' |
| Toy figurine of you | fun, gift-style | '1/7 scale figure on a desk, plastic texture, keep my face' |
| Cinematic couple shot | two people, moody | 'two people, keep both faces, cinematic dusk light' |
| Studio headshot | profile picture | 'plain backdrop, soft key light, natural skin' |
To get a usable set instead of one lucky image, lock your wording and reuse it. Keep the subject and keep line word for word and only change the scene or pose; the model holds likeness far better when most of the prompt stays put. When you need a different angle, add it plainly, like 'same person, same face, three quarter view, looking left.'
Results wobble run to run, so the second or third attempt is often the keeper. Save the wording that works so you can reuse it later.
Your style words are overpowering the likeness. Soften them and move 'keep my exact face' to the front so the model reads it first.
Add 'natural skin texture, visible pores' and start from a sharper input photo. Over-smoothing usually comes from a soft input, not the prompt.
Nano banana moderates hard on photos of real people, celebrities, and children, and tends to decline upfront. If you hit a wall, change the subject or use a photo you clearly have the right to use.
The prompts here are free to copy, and Gemini's free tier covers everyday edits; heavy batch use runs through the paid API at a few cents an image. Pick any prompt, drop in a sharp photo, and run it three times before you judge it.
Questions
It is the instruction you paste into Google's Gemini image tool, which people nickname nano banana. The model is Gemini 2.5 Flash Image. You give it a photo plus a line of text describing what to make and what to keep, and it returns an edited or generated picture. Banana is just the nickname, not part of the prompt.
Use five short pieces in plain English: the subject, what to keep from your photo, the look you want, the setting, and the framing. Put the keep line near the front so the face holds. Skip JSON or bracket formatting, it barely helps. One clear sentence beats a block of fake code.
Your style words are usually too strong, so the model rebuilds the face instead of keeping it. Move 'keep my exact face and skin tone' to the start of the prompt and ease off the heavy style language. Starting from a sharp, well lit photo helps a lot too. Run it a couple more times and pick the closest.
The prompts on this page are free to copy. Gemini has a free tier in the app that covers most casual edits. If you generate images in large batches you may hit limits or use the paid API, which runs a few cents per image. Try a prompt on the free plan first before paying for anything.
It moderates hard on certain things. Uploaded photos of real people, celebrities, and images of children get blocked often, and it tends to decline upfront rather than generate and then hide the result. If a prompt gets refused, change the subject or use a photo you clearly have the right to edit.
