Create layouts, refine selected regions, and combine visual elements with an image model built for design tasks and detailed editing.
Generate information graphics and design layouts that bring headings, imagery, and supporting details into one composition. Specify the reading order and essential content when preparing a poster, educational graphic, or product overview.
Supply the original image and identify the edit with a circled region, a box, an arrow, or coordinates. Describe the requested change, such as the color or material of a shoe’s upper. Review the result around the selected area and adjacent details.
Change an object’s color or material to develop a new visual direction. Supply a color value or a visual reference when needed, then inspect how the revised surface interacts with the scene’s lighting and surrounding elements.
Use a sketch to communicate arrangement before rendering a more detailed image. Rough lines and color blocks can indicate where elements belong, giving a poster, product concept, or scene a visual starting structure.
Combine multiple image references with instructions about their roles and placement. This supports visual collage and concept development, such as arranging supplied objects within a shared scene. Review the resulting scale, edges, and lighting together.
Generate visual designs with multilingual text, including languages with different writing directions and character forms. Provide the intended wording and layout requirements. Proofread every line and check reading order before preparing the design for publication.
| Dimension | Seedream 5.0 Pro | Nano Banana 2.1 | Nano Banana 2 Lite |
|---|---|---|---|
| Starting inputs | Text; up to 10 reference images, including sketches | Text and multiple reference images | Text and image |
| Core direction | Image generation, layout, and local editing | Reference-guided images and conversational editing | Image generation and multi-turn editing |
| Output type | Images | Text and images | Text and images |
| Aspect ratio options | Ark custom sizes: width-to-height ratio from 1:16 to 16:1 | Includes 1:4, 4:1, 1:8, and 8:1 panoramic formats | Includes 1:1, 9:16, 16:9, and 21:9 |
| Output resolution | Ark: 1K, 1.5K, 2K; default 2K | Gemini API: 1K, 2K, 4K | Gemini API: 1K; 2K and 4K not supported |
| Suggested uses | Layouts, localized visuals, and targeted image revisions | Reference-led concepts, text layouts, and image revisions | 1K image concepts and conversational revisions |
Specifications checked October 19, 2026. Sizes and controls refer to the named provider and task. Sources: Seedream 5.0 Pro model guide · Ark image generation API · Google Nano Banana 2.1 model card · Google Nano Banana 2 Lite model card
Plan the headline, subject and negative space as one composition. This suits campaign posters that combine imagery and text; check spelling and visual hierarchy before delivery.
Explore local color or material changes around an existing product image. This is useful for packaging and merchandise concepts; check that important structure and areas outside the requested edit remain suitable.
Use a sketch to establish composition and multiple images to suggest subjects, settings or style. This helps turn scattered ideas into a visual proposal; specify the role of each reference.
Explore image workflows for campaign layouts, product variations, and reference-led concepts.
Seedream 5.0 Pro is ByteDance’s image creation model for generation and design editing. Its documented capabilities include information layouts, region guided changes, material and color replacement, sketch rendering, multi image composition, and multilingual text. These features address different parts of a visual design brief.
The Ark API offers 1K, 1.5K and 2K generation, with 2K as the default, plus custom pixel dimensions. Custom width-to-height ratios range from 1:16 to 16:1 within the pixel-count limits. Image generation accepts up to 10 reference images.
Region guidance helps identify where a change should happen, but it is not a guarantee that every surrounding pixel will remain identical. Review the selected area and its neighbors after editing, especially when object structure, small text, edges, or lighting relationships need to stay consistent.
Yes. Multilingual text generation is a documented capability. Supply the intended wording and explain the layout requirements, including reading direction when relevant. Proofread the result in the target language and check whether the localized text fits the intended hierarchy without crowding other elements.
Pexo supports a conversational approach to creative work, including image generation and video ideas. Discuss the message, audience, and visual direction in plain language.
