OmwAIOmwAI
API GuideAPI ReferenceProgramming Tools
Seedream Image

Image Generation (Async)

Seedream asynchronous text-to-image and image-to-image tasks

POST /v1/images/async/generations submits a Seedream text-to-image or image-to-image task. Retrieve the result using the query endpoint. Unlike the synchronous Images API, this route uses the Volcengine image task adaptor.

Models and parameter scope

The adaptor declares doubao-seedream-4-0-250828 and seedream-4-0-250828; actual availability depends on channel configuration. Do not assume the synchronous Doubao plugin's 5.0 or 4.5 models, n, sequential generation or layer decomposition are available on this async route.

ParameterBehavior
modelRequired: one of the two declared names above; channels may map the upstream name.
promptRequired nonempty text for both text and reference-image generation.
imageOptional single reference image URL; enables image-to-image.
imagesOptional array of reference image URLs; takes priority over image. Image format and count limits depend on upstream.
sizeOptional, passed through to upstream; the adaptor does not specify preset sizes.
metadataOptional object merged into the upstream request; cannot change the mapped model. Use only fields supported by upstream.

This adaptor forwards references in the upstream image field; it does not accept GPT Image's top-level quality or n as documented parameters. Do not copy the synchronous parameter table to this endpoint.

Async workflow

  1. Submit and receive id / task_id and status: "queued" (submission acknowledgment only).
  2. Poll GET /v1/images/async/generations/{task_id}.
  3. When completed, read data[].url or upstream-provided data[].b64_json; for failed, inspect error.
POST
/v1/images/async/generations

Authorization

BearerAuth

AuthorizationBearer <token>

In: header

Request Body

application/json

model*string

火山引擎图片任务适配器注册的模型

Value in"doubao-seedream-4-0-250828" | "seedream-4-0-250828"
prompt*string

非空提示词;文生图和图生图都必须提供

image?string

单张参考图 URL;提供后进入图生图

images?array<string>

多张参考图 URL;提供后进入图生图

size?string

输出尺寸;任务适配器原样转发,具体取值依上游模型

metadata?

透传覆盖上游图片请求字段;不能借此更改 model,其他字段取值依上游

Response Body

application/json

curl -X POST "https://www.omwai.xyz/v1/images/async/generations" \  -H "Content-Type: application/json" \  -d '{    "model": "doubao-seedream-4-0-250828",    "prompt": "一只猫坐在窗边",    "size": "2K"  }'
{
  "id": "string",
  "task_id": "string",
  "status": "string",
  "model": "string",
  "created_at": 0
}
Empty

How is this guide?

Last updated on