Cat Selfie
4:3A selfie of a cat as the protagonist at twilight by the Eiffel Tower, happy, holding a piece of baklava, slight motion blur and slightly overexposed, with the text "Seedream 4.5 is on fal" at the top
Seedream 4.5
ID:
bytedance/seedream-4.5Seedream 4.5 image generation API by ByteDance — unified generation and editing in one model with up to 4K output and 14 reference images.
Text to imageImage editingReference images4K
Input
OutputView all
Output will appear here
basic
$0.05
high
$0.05
cURL
curl https://api.runbase.net/v1/runs \
-H "Authorization: Bearer $RUNBASE_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "bytedance/seedream-4.5",
"input": {
"prompt": "A cinematic product photo of a ceramic lamp",
"aspect_ratio": "1:1",
"resolution": "1K"
}
}'Examples
Art Exhibition Poster
3:4Minimalist art exhibition poster with an off-white background, a black line drawing of a human portrait slightly right of center, "A/W ART FAIR" in serif font at the top left and exhibition details below
Product Promo Layout
16:9A full-process cafe design tool promo for entrepreneurs and designers, covering store layout, functional zoning, decoration style, and equipment selection
Overview
Seedream 4.5 is ByteDance's image model released in December 2025. It unifies generation and editing into a single model — the same architecture handles both text-to-image and reference-guided editing without switching modes. It delivers strong text rendering, up to 4K resolution at the high quality setting (2K at basic), and accepts up to 14 reference images. It ranks #10 on LM Arena for image generation.
Use cases
Product photography and e-commerce assets with accurate text overlays. Editorial and campaign visuals where prompt adherence is critical. Poster and banner designs with embedded text. Reference-guided editing — upload up to 14 images (10 MB each) to steer style, composition, or content.
Inputs
All parameters are passed in the input object of the run request.
| Parameter | Required | Description |
|---|---|---|
| prompt | Yes | Text description (1–3000 chars) |
| aspect_ratio | No | Default 1:1. Options: 1:1, 2:3, 3:2, 3:4, 4:3, 9:16, 16:9, 21:9 |
| quality | No | Default basic (2K). Options: basic, high (4K) |
| image_urls | No | Up to 14 reference images (max 10 MB each) for editing |
Prompt tips
Be explicit about text placement
Seedream 4.5 handles embedded text well. Specify the exact string, position (e.g., "top left," "centered"), and font style for best results.
Use high quality for print-ready output
The basic setting outputs at 2K, which is fine for screens. Switch to high for 4K output when you need print-resolution or large-format imagery.
Leverage the 14-image reference limit
With support for up to 14 reference images at 10 MB each, you can provide extensive style guidance — mood boards, brand assets, and composition references in a single request.
Limitations
- Prompt max length is 3000 characters — shorter than many competitors
- 8 aspect ratios available (no ultra-wide options beyond 21:9)
- 4K output only available at
highquality setting - Maximum 14 reference images per request, 10 MB each
- Complex multi-element compositions may require prompt iteration
FAQ
What resolution does Seedream 4.5 output?
2K at basic quality (the default) and 4K at high quality. Both settings are the same price.
How many aspect ratios are supported?
Eight options: 1:1, 2:3, 3:2, 3:4, 4:3, 9:16, 16:9, and 21:9. The default is 1:1.
Can I edit existing images?
Yes. Upload up to 14 reference images via image_urls and describe the desired changes in the prompt. The model uses a unified architecture for both generation and editing.