Nano Banana
Nano Banana is Google's Gemini-based image generation and editing model, best known for keeping a subject looking like themselves across edits — change an outfit, hairstyle, or setting while the face stays consistent. It also does text-to-image from a plain description, and it runs right here on Civitai with no GPU or install.
About Nano Banana
Nano Banana is Google's image generation and editing model, delivered through the Gemini app's native image editing and hosted on Civitai. Google built it around a specific problem: when you edit a photo of yourself or someone you know, a result that is "close but not quite the same" doesn't feel right. So the model is tuned to maintain a character's likeness from one image to the next — you hand it a photo, describe the change you want, and it keeps the person (or pet) recognizably themselves while applying the edit.
Its standout is identity-preserving editing rather than raw text-to-image novelty. You can give a subject a costume or location change and keep their look consistent across every variation; blend multiple photos into one new scene (you plus your dog on a basketball court); edit in multiple turns, altering one part at a time — paint a room's walls, then add a bookshelf, then furniture — while the rest of the image is preserved; or apply the style of one image to an object in another, like mapping a butterfly's wing pattern onto a dress. It also generates images from a text prompt when you are starting from scratch.
Three versions are available on Civitai: the original Nano Banana, Nano Banana Pro, and Nano Banana 2. All are hosted (API-only) — there are no open weights to download and no local run — so you switch between versions in the generator without any setup. Choose Nano Banana when the job is editing or personalizing an existing photo with the subject kept intact; for open-weight photorealism, deep LoRA libraries, or fully local workflows, an ecosystem like Flux is the better fit.
How to prompt Nano Banana
- For edits, write the change as a plain instruction rather than a full scene description — "give her a 1960s beehive haircut" or "put a tutu on the dog." The model applies the change and keeps the subject looking like themselves, so you rarely need to re-describe the person.
- Provide a reference photo when likeness matters. Nano Banana is built to preserve a person's or pet's identity across edits, so upload the subject and describe only what should differ — the outfit, profession, decade, or location.
- Blend by supplying multiple images in one prompt. Give it two or more photos (for example you and your pet) and describe the combined scene; it composes them into a single coherent image instead of picking one.
- Edit in multiple turns for complex changes. Make one adjustment at a time and feed the result back in — paint the walls, then add a bookshelf, then a coffee table — so each step alters its target while preserving everything else.
- For text-to-image from scratch, describe the subject, setting, and lighting in natural sentences. For style transfer, name the source look and the target object explicitly — "apply the color and texture of these flower petals to a pair of rain boots."
AI models move fast — new versions ship often, and a model’s capabilities or Buzz cost can change. For the latest, check the model’s own page before you generate.
Featured Nano Banana models
Curated — the models worth generating with first.
Example generations
Curated, safe-for-work showcase — every image ships with its prompt and settings.
How to run Nano Banana
Nano Banana is a hosted API model — skip the setup and generate on Civitai.
⚡ Run on Civitai (Recommended)
The fastest way to start.
🔌 API-only model
Nano Bananaruns through its provider's API — there are no public weights to download and nothing to install.
Civitai handles the API access — you just prompt and generate.
Nano Banana vs other ecosystems
Backed by Civitai usage data.
| Feature | Nano Banana | Imagen 4 | Seedream | Flux |
|---|---|---|---|---|
| Best for | Likeness-preserving photo editing | Photoreal text-to-image | High-res text-to-image | Photorealism, text, versatility |
| Image editing | Native, multi-turn | Limited | Via edit variant | In-context (Kontext) |
| Character consistency | Excellent | Fair | Good | Fair |
| Text-to-image quality | Very good | Excellent | Excellent | Excellent |
| Access | API (hosted) | API (hosted) | API (hosted) | Open weights + API |
| Available on Civitai | ✓ Yes | ✓ Yes | ✓ Yes | ✓ Yes |
Frequently asked questions
How much does it cost to generate with Nano Banana?
What makes Nano Banana different from other image models?
What's the difference between Nano Banana, Nano Banana Pro, and Nano Banana 2?
Can it keep a person or pet looking consistent across edits?
Can I use LoRAs with Nano Banana?
Do I need a GPU to run Nano Banana?
Start generating with Nano Banana now
No installation. No GPU. Runs in your browser.