Prompt-generated images include the Wan 3.0 site mark when removal is not available on your account.
View paid options for watermark control
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.
Each block is a different job. Pick the one that matches the still you need.
01Useful 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
02Slogans, 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
03One 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
04Follow-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
05Native 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 sizeStart from a prompt or an uploaded still, then generate or edit toward one visual target.
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.
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.
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.
Match the task to what each model is for. These are workflow differences, not scores.
| Capability | Nano Banana Pro | Nano Banana 2 | Seedream 5.0 Pro |
|---|---|---|---|
| Infographics | Diagrams, explainers, current facts | Fast drafts, optional current context | Dense infographics and campaign boards |
| In-image text | Multilingual posters, mockups, paragraphs | Localized layouts and campaign type | Dense design type and layered copy |
| References | Up to 14 stills, 5 people | Up to 10 objects, 4 characters | Region-aware fusion and identity hold |
| Studio edits | Lighting, camera, focus, local regions | Conversational edits and multi-turn refine | Region-aware layers and object restyle |
| Output | Native 1K, 2K, 4K | 0.5K through 4K, including ultra-wide | Campaign stills and layered comps |
Open another model page when you need a different mix of speed, layout, or editing.
Create a still, revise a source, or prepare a frame for video.
Generation, editing, in-image text, and how this page works.
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.
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.
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.
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.
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.
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.
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.
Prompt-generated images include the Wan 3.0 site mark when removal is not available on your account.
View paid options for watermark controlUse 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.
Each block is a different job. Pick the one that matches the still you need.
01Useful 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 prompt02Slogans, 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 image03One 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 editing04Follow-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 source05Native 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 sizeStart from a prompt or an uploaded still, then generate or edit toward one visual target.
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.
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.
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.
Match the task to what each model is for. These are workflow differences, not scores.
| Capability | Nano Banana Pro | Nano Banana 2 | Seedream 5.0 Pro |
|---|---|---|---|
| Infographics | Diagrams, explainers, current facts | Fast drafts, optional current context | Dense infographics and campaign boards |
| In-image text | Multilingual posters, mockups, paragraphs | Localized layouts and campaign type | Dense design type and layered copy |
| References | Up to 14 stills, 5 people | Up to 10 objects, 4 characters | Region-aware fusion and identity hold |
| Studio edits | Lighting, camera, focus, local regions | Conversational edits and multi-turn refine | Region-aware layers and object restyle |
| Output | Native 1K, 2K, 4K | 0.5K through 4K, including ultra-wide | Campaign stills and layered comps |
Open another model page when you need a different mix of speed, layout, or editing.
Create a still, revise a source, or prepare a frame for video.
Generation, editing, in-image text, and how this page works.
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.
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.
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.
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.
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.
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.
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.
