Latest Nano Banana model
What is Nano Banana 2?
Nano Banana 2 is Google's Gemini 3.1 Flash Image model for fast image generation and editing with Pro-level visual quality. This release improves text rendering, subject consistency, output control up to 4K, and workflows that use as many as 14 reference images. It is distinct from the older Pro API while aiming for professional results with a faster Flash architecture.
Generation modes and practical strengths
Flash-speed creative iteration
Use the original model when cost and quick drafts matter most. Choose Nano Banana 2 when a production concept needs more detail, sharper typography, or a higher-resolution deliverable.
Text and layout control
The newer model is designed to render clearer words inside posters, packaging concepts, interface mockups, and localized campaign visuals. Always proofread generated text before publishing.
Multi-image generation and editing
Add up to 14 references to guide identity, products, composition, or style. A focused prompt should explain the role of each image and the exact change the output needs.
A repeatable image workflow
Start with one clear visual objective, select the model that fits the required quality and budget, then refine one variable at a time.
- 1Choose the 4-credit original for a quick draft or version 2 for higher-resolution, reference-led work.
- 2Write a prompt that specifies the subject, composition, lighting, camera feel, style, and any text that must appear.
- 3For version 2 edits, add only useful reference images and describe what each one should influence.
- 4Review anatomy, typography, product details, and consistency, then change one prompt detail or output setting before regenerating.
Nano Banana 2 vs GPT Image 2 vs Seedream 5.0 Pro
The best image model depends on the job. Compare workflow fit instead of treating one benchmark or sample as universal.
Nano Banana 2
Best fit here for fast Pro-level generation, readable text, up to 4K output, and edits guided by multiple reference images.
GPT Image 2
A useful comparison choice for instruction-led image creation and editing when an OpenAI-centered workflow is more important than this page's live Kie controls.
Seedream 5.0 Pro
A relevant alternative for polished commercial visuals and complex creative direction; compare its available controls and price before choosing a production workflow.
Prompt patterns you can reuse
For Nano Banana 2, a useful prompt separates the subject, visual goal, composition, lighting, typography, and constraints. Start with the outcome you need, then add only details that help the model make a visible decision.
Product campaign prompt
Create a premium studio photograph of a matte black travel bottle on pale stone. Use soft morning side light, a subtle reflection, a warm neutral background, and generous negative space on the left for campaign copy. Keep the logo shape, cap proportions, and material finish consistent with the reference image. Do not add extra objects, labels, or decorative text.
Poster with readable text
Design a vertical launch poster for a summer skincare collection. Place the headline “Light on skin” in large clean sans-serif type, followed by “Daily mineral protection” as a smaller line. Use coral and cream blocks, one centered product, crisp spacing, and a print-ready editorial layout. Preserve every supplied word exactly and avoid invented small print.
Reference-led edit prompt
Use reference one for the person and clothing, reference two for the room layout, and reference three only for the color palette. Move the subject beside the window, change the scene to late-afternoon light, and keep facial identity, garment details, furniture scale, and camera perspective unchanged. Remove the table clutter without replacing it with new props.
Review before you publish
Treat the first result as a draft. Check objective details before judging style, then revise one variable at a time so you know which instruction improved the next generation.
Proofread every visible word and verify logos, packaging, hands, faces, reflections, and small product geometry at full size. Compare reference-led outputs against each source image, confirm that the selected aspect ratio and resolution fit the delivery channel, and check that edits did not introduce unrequested objects or remove important details. For a revision, name the single problem, state what must remain unchanged, and regenerate with the same model and settings before changing anything else.
Credit pricing by resolution
The workspace shows the charge before submission: 4 credits for the original model, or 30/45/68 credits for version 2 at 1K/2K/4K. Failed provider tasks are refunded through the existing task-credit flow.
Image generation FAQ
What is Nano Banana 2?
It is Google's Gemini 3.1 Flash Image model, combining fast generation with Pro-level image quality, clearer text, multi-image references, and output up to 4K.
Is Nano Banana 2 the same API as Nano Banana Pro?
No. Kie identifies Nano Banana 2 as Gemini 3.1 Flash Image and the older Pro release as Gemini 3 Pro Image. This version targets Pro-level quality with a faster Flash architecture.
How many credits does Nano Banana use?
The original uses 4 credits per image. Version 2 uses 30 credits for 1K, 45 for 2K, or 68 for 4K output.
Can I edit an existing image?
Yes. Select version 2, upload up to 14 reference images, and explain what to preserve, replace, combine, or restyle in the prompt.