Nano Banana 2 lets you choose aspect ratio and resolution when generating or editing images through Google's Gemini API. The Google DeepMind model is officially called Gemini 3.1 Flash Image. Access is through Google services, with no open weights available. Launch and model documentation.
What are the key facts about Nano Banana 2?
| Developer | Released | Type | Size or parameters | License and access | Where it runs |
|---|---|---|---|---|---|
| Google DeepMind. Launch | February 26, 2026. Launch | Image generation and conversational editing. Model page | Not published in the model documentation. Model page | Hosted service under Google's API terms; no open weights. Terms and access | Google's infrastructure through Gemini API and AI Studio. Access |
Three decisions belong in an image-format brief: the shape of the canvas, the resolution tier, and the composition inside that canvas. Write these separately. An instruction such as “make a banner” leaves both the intended shape and the placement of the subject open to interpretation.
For the product's identity and release history, see the sibling Nano Banana 2 release and feature guide.
Which image formats and ratios does Nano Banana 2 support?
Google documents additional wide and tall aspect ratios for Nano Banana 2, including 4:1, 1:4, 8:1, and 1:8. Its model page also lists resolution choices from 0.5K through 4K. These controls make it possible to ask for a composition intended for a particular placement. Release details and model capabilities.
For example, plan a product campaign around a landscape image and a portrait image. In the landscape brief, request open space beside the product for a later headline. In the portrait brief, request space above it. Judge each result against its own layout rather than treating one as an automatic replacement for the other.
The API guide exposes aspect ratio and image size as explicit configuration fields. Prefer setting them there when building an application, and use the prompt to describe where the subject belongs. That gives your saved request an inspectable format choice and a separate creative instruction. Output configuration.
What should you check before approving a reformatted image?
Set the aspect ratio and resolution tier independently: Google's API exposes aspect_ratio and image_size as separate fields. Inspect the returned image's dimensions before delivery, and record them alongside the requested settings. Output configuration.
Treat preservation as a review requirement: compare the reference and output wherever exact details matter. Check the product outline, label, and important background elements individually before approving a new composition.
Google also notes that the model may not return the exact number of images requested. Ask for one deliverable at a time during initial testing, inspect what comes back, and handle missing outputs explicitly in your application. Known limitations.
A format change can be a design decision as well as an API setting. Before approving a wider image, check whether the product still has the intended prominence. For a tall image, check whether the added space serves the layout rather than merely making the canvas longer.
How do you set Nano Banana 2 aspect ratio and resolution?
Create a Gemini API key using Google's documented setup and make it available as GEMINI_API_KEY. Select gemini-3.1-flash-image for this example. The following REST request uses the current image-generation guide's Interactions API and its response_format configuration. Key setup and API example.
curl --fail-with-body \
https://generativelanguage.googleapis.com/v1beta/interactions \
-H "x-goog-api-key: $GEMINI_API_KEY" \
-H 'Content-Type: application/json' \
-d '{
"model": "gemini-3.1-flash-image",
"input": "Create a ceramic lamp on a desk, on the right, with empty wall space on the left.",
"response_format": {
"type": "image",
"aspect_ratio": "16:9",
"image_size": "2K"
}
}'
This returns an API response containing generated image data, rather than writing a picture directly to disk. Google's guide shows how the SDK exposes the generated image and how to decode and save its data. Use that response handling when incorporating the request into an application. Saving generated images.
For a controlled format experiment, start with one composition. Keep the prompt and output tier recorded, generate the landscape version, and inspect the subject's position. Then request the portrait layout with an instruction suited to that placement. These are separate creative requests, so review them separately.
For an existing photograph, follow the guide's image-editing example and supply the reference image alongside your text. Describe what should remain recognizable and what space should be added or rearranged. Avoid piling a new location, new lighting, a new product color, and a new aspect ratio into the first revision. Reference-image input.
Use a delivery checklist with concrete questions: Is the whole subject visible? Is the intended text area empty? Does the background reach every edge? Does the image still communicate the same message when viewed in its intended placement? Save the approved version together with the format request.
If the composition is suitable but the destination requires a different final file specification, finish the delivery as a separate step. Record the generated dimensions and the delivered dimensions independently. This keeps your generation settings understandable to the next person working on the asset.
How does Nano Banana 2 compare with Pro for image formats?
Nano Banana Pro is a real alternative for image creation, but format support should be checked per model. Google's Nano Banana 2 release specifically identifies the added extreme aspect ratios, while Pro's model page emphasizes graphic design, product mockups, and search-grounded visuals. Choose a candidate based on the required controls, then test your composition. Nano Banana 2 formats and Pro capabilities.
| Requirement | What to evaluate |
|---|---|
| Very wide or tall layout | Nano Banana 2's documented additional ratios. Launch |
| Complex branded composition | Nano Banana Pro's documented design capabilities. Model page |
Use the ComfyUI guide when planning how an image will move through a larger workflow. Keep format requirements attached to the asset brief so they remain clear whichever interface you use.
What else should you know about Nano Banana 2 image formats?
Is Format Image Nano Banana 2 a separate model?
No: Google's model is Nano Banana 2, or Gemini 3.1 Flash Image. Image formatting here refers to selecting output proportions and resolution. Model page.
Can Nano Banana 2 make portrait and landscape images?
Yes: Google documents multiple aspect ratios, including 9:16 and 16:9. Select the ratio explicitly and describe the composition you need inside it. Aspect-ratio configuration.
How do I request a 2K image with Nano Banana 2?
Set image_size to "2K" in the image response_format, and set aspect_ratio separately. Inspect the returned file before approving it for a placement with exact pixel requirements. Output configuration.
Can I run this formatting model offline?
Nano Banana 2 is a hosted Gemini model with no open weights. Changing image-format settings does not change that access model. Developer access.
Sources
Primary sources checked on September 5, 2026.
- Google developer announcement of Nano Banana 2
- Google model documentation for Gemini 3.1 Flash Image
- Google Gemini API terms
- Google Gemini API image-generation documentation
- Google documentation for Gemini API keys
- Google model documentation for Gemini 3 Pro Image
Top comments (0)