1 available
0 / 20000

Prompt-generated images include the Wan 3.0 site mark when removal is not available on your account.

View paid options for watermark control
Cost 3 credits / Sign in to view your creditsBuy Credits ›
Museum atrium wayfinding totem photographed in documentary light
Keep this tab open until the prompt task finishes; the completed result is also recorded in My Creations.

Nano Banana Pro image generation and editing guide

Use this model for infographics, product mockups, posters with real copy, and scenes assembled from several references. It can place multilingual type in the frame, keep up to five people consistent, mix as many as 14 reference stills, and output 1K, 2K, or 4K.

Use this guide to define the prompt, aspect ratio, and resolution, then test the brief in the current Wan 3.0 workspace with an available image model.

Open the current workspace
Foggy cedar forest boardwalk at dawn

What it is built to handle

Each block is a different job. Pick the one that matches the still you need.

Botanist workbench with plant specimens, vials, and handwritten labels
01
01

Infographics and labeled diagrams

Useful when the still has to carry steps, labels, or current facts. Weather, places, or timely details can inform the frame when current-context lookup is on. Supply notes you already trust, and check every number and proper name at full size.

Draft the prompt
Print technician inspecting a civic poster proof with a short fictional festival headline
02
02

Multilingual in-image text

Slogans, paragraphs, mockups, menus, and packaging copy can sit inside the composition, including localized languages. Paste the final wording. Check kerning and line breaks at output size; rebuild prices and legal lines in a design file before publishing.

Open text to image
Five adults photographed together in a sunlit loft
03
03

Multi-reference consistency

One composition can mix up to 14 reference stills, with resemblance held for up to five people, plus object and style references. Label each file—person, product, material, or look. Conflicting wardrobe, lens, or lighting across those files is a common reason identity drifts.

Open reference editing
Closed cafe table at night with hard lamp light and rain on the window
04
04

Lighting, camera, and local edits

Follow-up instructions can change lighting, camera height, focus, color grade, or a selected region while the rest of the frame stays put. Ask for one change at a time: day to night, a tighter crop, or a relit product. Then check contact shadows and whether protected faces or packaging still match.

Edit from a source
Macro still of an unbranded mechanical watch movement under raking light
05
05

1K to 4K output

Native 1K, 2K, and 4K cover a first direction through a large still. Set aspect ratio to the channel before you write a long prompt. Wide ratios such as 21:9 are available on the model; this workspace offers the common ratios used on Wan 3.0. Check texture, type, and edges at the size you will ship.

Set output size

How to use it

Start from a prompt or an uploaded still, then generate or edit toward one visual target.

01

Choose generation or editing

Use text-to-image for a new composition. Use image-to-image when a source must stay recognizable. Set aspect ratio and 1K, 2K, or 4K before you write a long prompt.

02

Describe the result

Name the subject, scene, required wording, language, and details that must not change. For an edit, write the change first and the preservation list second. Label each reference if you attach more than one file.

03

Generate and refine

Review the first still at a usable size, then correct type, light, crop, or a protected region in one pass. Keep the accepted version with its prompt so later turns stay comparable.

Open the current image workspace

Compare with other image models

Match the task to what each model is for. These are workflow differences, not scores.

CapabilityNano Banana ProNano Banana 2Seedream 5.0 Pro
InfographicsDiagrams, explainers, current factsFast drafts, optional current contextDense infographics and campaign boards
In-image textMultilingual posters, mockups, paragraphsLocalized layouts and campaign typeDense design type and layered copy
ReferencesUp to 14 stills, 5 peopleUp to 10 objects, 4 charactersRegion-aware fusion and identity hold
Studio editsLighting, camera, focus, local regionsConversational edits and multi-turn refineRegion-aware layers and object restyle
OutputNative 1K, 2K, 4K0.5K through 4K, including ultra-wideCampaign stills and layered comps

Frequently asked questions

Generation, editing, in-image text, and how this page works.

What is Nano Banana Pro?

An image generation and editing model. It takes text and images and returns stills. Typical work includes infographics, multilingual type in the frame, multi-reference scenes, lighting and camera edits, and native 1K to 4K output.

What is Nano Banana Pro good at?

Jobs where layout, type, or people have to stay stable: infographics, product mockups, posters with real wording, localized packaging, and group or product scenes built from several references. Use a faster sibling model when you only need many cheap directions.

Can Nano Banana Pro edit an existing image?

Yes. Upload a source, state the change, and list what must remain: identity, product geometry, crop, lighting, or label area. Check unchanged regions as carefully as the new area.

Can Nano Banana Pro render readable text in an image?

Yes. Provide the exact wording and language, then check every character, number, and line break at full size. Rebuild prices and legal copy in a design file before publishing.

Can I try Nano Banana Pro on this page?

Not directly. Nano Banana Pro is not in the current model selector. Use this page to prepare the brief, then test it in the Wan 3.0 workspace with an available image model.

Which Nano Banana Pro inputs and outputs are documented?

The model accepts text and image inputs for generation and editing. Documented controls include prompts, reference images, aspect ratio, and 1K, 2K, or 4K output.

Carry the brief into the current Wan 3.0 workflow

Lock channel, ratio, protected marks, and the one change you need. Test the brief with an available image model, then keep the accepted still with its prompt for the next crop, language version, or image-to-video test.

Radio telescope dish in a high desert at night