Flagship output quality
Seedream 5 Pro renders photoreal scenes, editorial product shots, and accurate in-image text at 1K or 2K — competitive with GPT Image and NanoBanana tiers.
Seedream 5 Pro API guide
Seedream 5 Pro is ByteDance's flagship image model — strong multilingual text rendering, multi-reference editing, and clean photoreal output. On Seedance API you call it as a stable alias with simple REST endpoints, async task polling, and per-image credit pricing.
1K or 2K output
1:1 to 16:9, 9:16, 2:3, 3:2
Prompt or up to 10 reference images
JSON over HTTPS
Starter request
curl -X POST 'https://seedanceapi.org/v1/images/generations' \
-H 'Authorization: Bearer YOUR_API_KEY' \
-H 'Content-Type: application/json' \
-d '{
"model": "seedream-5-pro",
"prompt": "Studio product shot of a matte black water bottle on wet stone, soft top light",
"aspect_ratio": "4:3",
"resolution": "2k",
"image_urls": []
}'POST /v1/images/generations returns a task_id. Poll GET /v1/images/generations/:task_id until status is SUCCESS, then read images[].url.
Teams pick the Seedream 5 Pro API when they need a flagship-quality image model behind a boring, predictable REST surface — no provider SDK, no surprise schema changes.
Seedream 5 Pro renders photoreal scenes, editorial product shots, and accurate in-image text at 1K or 2K — competitive with GPT Image and NanoBanana tiers.
You call model "seedream-5-pro". Provider routing, capacity, and fallbacks stay on our side — your integration never touches upstream changes.
9 credits per 1K image, 18 per 2K, plus 0.8 per reference image after the first. estimated_credits comes back in every create response.
The public surface exposes what the model actually does: text-to-image generation and reference-guided editing with up to ten input images.
Send a prompt up to 3000 characters with an aspect ratio and resolution. The model handles composition, lighting, and in-image typography.
Pass 1–10 public HTTPS image URLs in image_urls to edit, composite, or restyle existing visuals — product swaps, background changes, style transfer.
Set resolution to "2k" for print-quality renders, marketing assets, and detailed composites. 1K covers preview and iteration loops.
One of Seedream 5 Pro's strongest capabilities — readable English, Chinese, Japanese, and mixed-language copy inside posters and signage.
Creation returns a task_id immediately. Poll status, retry on your schedule, and keep your UI responsive — no long-held HTTP connections.
Response shape stays stable regardless of upstream routing. consumed_credits, status, and image URLs are normalized before they reach you.
Three steps from API key to rendered image URL.
Sign in, open the API Keys page, and copy a key for server-side use. Keep it on your backend — never ship it in browser code.
Call POST /v1/images/generations with model "seedream-5-pro", your prompt, aspect_ratio, resolution, and optional image_urls. You get back a task_id and estimated_credits.
Call GET /v1/images/generations/:task_id. When status reaches SUCCESS, read images[0].url and display or archive the result.
Two endpoints cover the full lifecycle — create and poll.
Creates a Seedream 5 Pro task. Empty image_urls means text-to-image; 1–10 URLs switches to reference editing. Returns 202 with task_id and estimated_credits.
Returns IN_PROGRESS, SUCCESS with images[].url, or FAILED with an error object. Also reports consumed_credits once the job finishes.
400 invalid params, 401 bad key, 402 insufficient credits, 404 unknown task, 429 rate limit, 503 upstream unavailable. Full table in the docs.
| Parameter | Value |
|---|---|
| model | Must be "seedream-5-pro" — the only public model on this route. |
| prompt | Text prompt or editing instruction. 1–3000 characters. |
| aspect_ratio | One of 1:1, 4:3, 3:4, 16:9, 9:16, 2:3, 3:2. Default 1:1. |
| resolution | "1k" (default, 9 credits) or "2k" (18 credits). |
| image_urls | 0–10 public HTTPS URLs. Empty array = pure text-to-image. |
Unknown parameters are rejected with 400 UNKNOWN_PARAMETER — the contract stays explicit. Full schema in /docs/images.
Flat per-image pricing, billed in Seedance API credits. The create response includes estimated_credits; the finished status reports consumed_credits.
| Mode | Credits | Notes |
|---|---|---|
| Text-to-image · 1K | 9 credits / image | Default resolution. Best for iteration loops and previews. |
| Text-to-image · 2K | 18 credits / image | Print-quality output for final assets. |
| Reference editing | +0.8 credits / extra image | First reference image included in base price; images 2–10 add 0.8 each. |
Answers for developers evaluating the Seedream 5 Pro API on Seedance API.
Seedance API exposes Seedream 5 Pro behind a stable REST surface — you get the model's full capability with normalized auth, billing, and responses, without managing provider credentials yourself.
Send an empty image_urls array for text-to-image. Pass 1–10 public HTTPS image URLs to edit or composite references — the same endpoint handles both modes.
1:1, 4:3, 3:4, 16:9, 9:16, 2:3, and 3:2. Pick the ratio that matches your placement — the model composes for it natively rather than cropping.
9 credits per 1K image and 18 per 2K. Reference editing adds 0.8 credits for every image after the first. The create response returns estimated_credits so you can confirm cost before polling.
Typical 1K tasks finish in roughly 15–25 seconds; 2K and multi-reference edits take longer. Poll the status endpoint — tasks are async by design.
No — and you shouldn't. Keep your Bearer key on your backend, call our endpoints server-side, and expose only your own safe routes to clients.
Yes — multilingual in-image text is one of its strengths. Posters, price signs, and localized ad creative come out with readable copy in English, Chinese, Japanese, and more.
The complete request schema, response examples, error table, and pricing notes live in /docs/images. This page is the integration overview.
Get an API key, POST your first generation task, and plug flagship image generation into your product today.