Omni-Image API

Omni-Image is the OmniRouters image generation entry point for text-to-image, image-to-image, GPT image generation, async image tasks, multi-image generation, and panorama generation.

This page follows section 2.3 Minimal Request Example from the Feishu integration document. The examples below are the current working shapes; legacy model limits and old parameter notes are not duplicated here.

Routes

Method	Path	Description
POST	https://omnirouters.com/v1/images/generations	Create an image task synchronously or return image results directly
POST	https://omnirouters.com/v1/images/generations?async=true	Create an async image task
GET	https://omnirouters.com/v1/images/generations/{task_id}	Query an async image task

For Gemini-native contents / parts payloads, see Gemini Generate Content. They are not mixed into this standard Omni-Image page.

Authentication

Authorization: Bearer <your-api-key>
Content-Type: application/json

Usage

For synchronous creation, send the request body for the target workflow to POST /v1/images/generations. For async creation, add ?async=true to the same route, then query the returned task_id with the lookup route.

For async polling, a 5-second or 10-second interval is recommended.

Minimal Request Body Examples

Text-to-Image

{
  "model": "Nano-Banana2",
  "resolution": "4K",
  "prompt": "一个女孩在海边回头微笑，电影感光影，高清细节",
  "aspect_ratio": "16:9"
}

Image-to-Image

{
  "model": "Nano-Banana2",
  "resolution": "4K",
  "prompt": "将图1的服装换为图2的服装，整体自然协调，时尚杂志风格",
  "images": [
    "https://example.com/image1.png",
    "https://example.com/image2.png"
  ],
  "aspect_ratio": "16:9"
}

GPT Image Generation

{
  "model": "GPT-image-2",
  "quality": "low",
  "resolution": "4K",
  "prompt": "将图1的服装换为图2的服装，整体自然协调，时尚杂志风格",
  "images": [
    "https://example.com/image1.png",
    "https://example.com/image2.png"
  ],
  "aspect_ratio": "16:9"
}

Async Image Task

Create the task:

curl --location --request POST 'https://omnirouters.com/v1/images/generations?async=true' \
  --header 'Authorization: Bearer sk-xxxx' \
  --header 'Content-Type: application/json' \
  --data-raw '{
    "model": "Nano-Banana2",
    "resolution": "4K",
    "prompt": "将图1的服装换为图2的服装，整体自然协调，时尚杂志风格",
    "images": [
      "https://example.com/image1.png",
      "https://example.com/image2.png"
    ],
    "aspect_ratio": "16:9"
  }'

Poll the result:

curl --location --request GET 'https://omnirouters.com/v1/images/generations/{task_id}' \
  --header 'Authorization: Bearer sk-xxxx'

Reference URL Region

{
  "model": "Nano-Banana2",
  "resolution": "4K",
  "prompt": "将图1的服装换为图2的服装，整体自然协调，时尚杂志风格",
  "images": [
    "https://example.com/image1.png",
    "https://example.com/image2.png"
  ],
  "aspect_ratio": "16:9",
  "input_region": "Mainland"
}

input_region marks the region of the input reference image URLs:

Value	Description
Mainland	Mainland China
Oversea	Overseas
OverseaUSWest	Overseas, US West

Kling Generate Multiple Images

{
  "model": "Kling:3.0-omni",
  "resolution": "4K",
  "prompt": "一只小猫坐在窗边",
  "aspect_ratio": "16:9",
  "n": 3
}

Seedream Generate Multiple Images

{
  "model": "Seedream:4.5",
  "resolution": "4K",
  "prompt": "一只小猫坐在窗边",
  "n": 3
}

GPT-image-2 Generate Multiple Images

{
  "model": "GPT-image-2",
  "quality": "low",
  "resolution": "4K",
  "prompt": "将图1的服装换为图2的服装，整体自然协调，时尚杂志风格",
  "images": [
    "https://example.com/image1.png",
    "https://example.com/image2.png"
  ],
  "aspect_ratio": "16:9",
  "n": 3
}

Hunyuan World Panorama 2.0

{
  "model": "Hunyuan-3d-2.0",
  "prompt": "一个雪山湖泊的360度全景图，真实摄影风格"
}

Request Body

These are the common top-level fields that appear in the section 2.3 examples. For each workflow, send only the fields shown in its example.

Field	Type	Description
model	string	Image model name.
prompt	string	Image generation prompt.
resolution	string	Image resolution or size tier, for example 4K.
aspect_ratio	string	Image aspect ratio, for example 16:9.
images	string[]	Reference image URL array.
quality	string	Quality tier for GPT image workflows, for example low.
n	number	Number of images to generate in one request.
input_region	string	Region of input reference image URLs, for example Mainland.

Async creation uses the query parameter async=true; it is not a request body field.

Query an Async Task

curl https://omnirouters.com/v1/images/generations/$TASK_ID \
  -H "Authorization: Bearer $OMNIROUTERS_API_KEY"

Response Body

The endpoint usually returns an image result object or an async task object. Common fields are listed below; use the actual response as the source of truth:

Field	Type	Description
data	array	Image result array.
data[].url	string	Image URL.
data[].b64_json	string	Base64 image content.
id / task_id	string	Async task id.
status	string	Async task status.
error / message	string or object	Failure reason or error details.

Related Links

• API Reference overview
• Omni-Video API
• Gemini Generate Content
• Full Apifox reference