A colossal chrome whale breaching through luminous storm clouds above a submerged futuristic coral city, cascading water, dramatic volumetric light, surreal cinematic realism, no text, logos, or watermarks.
Seedream 4.5 Text to image API
Seedream 4.5 combines text-to-image creation with reference-guided editing for campaign visuals, product imagery, posters, characters, and multi-image compositions.
The playground matches the public request shape: prompt, image_urls for image editing, and aspect_ratio.
Pricing
Image pricing
- Generated image
- $0.04
Playground Examples
Try production-style prompt starters.
An editorial portrait wearing an intricate obsidian glass and molten-gold sculptural headdress, deep crimson background, bold chiaroscuro, museum-quality couture photography, no text, logos, or watermarks.
A monumental brutalist residence bridging volcanic cliffs above a turquoise waterfall dropping into a stormy ocean, warm glass interiors, dramatic aerial photography, no text, logos, or watermarks.
Similar Models
Compare adjacent capabilities before switching models.
ByteDance · Image
Seedream V5 Pro
ByteDanceByteDance image generation and editing for cinematic visuals, product shots, precise local edits, and reference-guided creative work.
OnlineTry model
Google · Image
Nano Banana Pro
GoogleGoogle's premium image generation and editing model for readable text, product mockups, visual explainers, and reference-guided brand assets.
OnlineTry model
Google · Image
Nano Banana 2
GoogleGoogle image generation and editing for photorealistic outputs, multi-reference guidance, and 1K to 4K results.
OnlineTry model
Model overview
Seedream 4.5 for image generation and reference-guided editing
Seedream 4.5 brings ByteDance image generation and editing to sjolt. Start with a prompt or combine up to 10 reference images for consistent subjects, creative compositions, product visuals, posters, and typography-heavy designs.
Create from text or visual references
Generate a new image from a written prompt, or use reference images to preserve subjects, products, materials, lighting, and visual direction.
- Supports Text to image and reference-asset generation workflows.
- Fits product assets, ad previews, visual direction, and social content tests.
- Adjust configured inputs, media, and switches before generation.
Compose with multiple images
Combine elements from up to 10 source images and direct the result with natural-language instructions for controlled creative edits.
Prompt-led generation
Build polished layouts and typography
Create posters, campaign graphics, product visuals, and editorial compositions with stronger layout and text rendering.
- Compare available models from Google, OpenAI, ByteDance, Kuaishou in one place.
- Copy the request body directly to the server to reduce frontend/backend parameter drift.
- Failure states, retries, result preview, and review checkpoints stay in the same workflow.
Model characteristics
Built for consistent subjects, multi-image composition, and production-ready visual direction.
Reference consistency
Preserve recognizable subject details, lighting, color, and materials while changing the surrounding composition.
Natural-language editing
Describe replacements, removals, layout changes, and scene transformations without masks or layer controls.
Flexible framing
Generate auto, landscape, portrait, square, 4:3, or 3:4 output for common creative formats.
FAQ
These are the first questions to answer when evaluating this model.
Can Seedream 4.5 generate without a reference image?
Yes. Use the text-to-image variant with a prompt and optional aspect_ratio.
How many reference images can I use?
The image-edit variant accepts up to 10 image_urls.
Which aspect ratios are available?
sjolt exposes auto, 16:9, 1:1, 9:16, 4:3, and 3:4. auto is kept local and is not forwarded to Comfy.