Images

OpenAI Images-format text-to-image and image editing endpoints. Available models (DALL·E 3, GPT-Image, Seedream, etc.) vary by channel — the fields below are the common ones.

Generate an Image

POST /v1/images/generations

RequestREQUEST

POSThttps://test.kailin.com.cn/v1/images/generations

The returned `url` is temporary — re-host promptly. b64_json mode returns large payloads and may render slowly.

API KeyStored only in your browser, never uploaded

Body (JSON){ "model": "dall-e-3", "prompt": "A cyberpunk city at night, neon lights, reflections after rain", "size": "1024x1024" }

ResponseRESPONSE

Fill in the request and hit Send — the response shows up here

Request

curl https://test.kailin.com.cn/v1/images/generations \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer sk-your-token" \
  -d '{
    "model": "dall-e-3",
    "prompt": "A cyberpunk city at night, neon lights, reflections after rain",
    "n": 1,
    "size": "1024x1792"
  }'

Request Body

Parameter	Type	Required	Description
model	string	No	Defaults to dall-e-2; also dall-e-3, gpt-image-1, etc.
prompt	string	Yes	Image description — up to 4000 characters on DALL·E 3
n	integer	No	Number of images 1–10 (DALL·E 3 only supports 1)
size	string	No	Resolution: 1024x1024 / 1792x1024 / 1024x1792 (DALL·E 3); 256/512/1024x1024 (DALL·E 2); auto/1024x1536/1536x1024 (GPT-Image)
quality	string	No	standard / hd (DALL·E 3); low/medium/high/auto (GPT-Image)
response_format	string	No	url (default) or b64_json (GPT-Image always returns b64)
style	string	No	DALL·E 3 only: vivid / natural
seed	integer	No	Random seed

Response

{
  "created": 1730000000,
  "data": [
    {
      "url": "https://oss.example.com/img/abc.png",
      "revised_prompt": "A cyberpunk city at night…"
    }
  ]
}

With response_format: "b64_json", data[].b64_json holds the Base64 image data. url results point to a temporary upstream location — download and re-host them promptly.

Edit an Image

POST /v1/images/edits

Upload the source image as multipart/form-data and modify it via a prompt (supported by GPT-Image / DALL·E 2):

curl https://test.kailin.com.cn/v1/images/edits \
  -H "Authorization: Bearer sk-your-token" \
  -F "model=gpt-image-1" \
  -F "prompt=Replace the background with a seaside at dusk" \
  -F "image=@/path/to/input.png"

Field	Type	Required	Description
image	file	Yes	Source image (png/jpeg/webp; GPT-Image accepts multiple image[])
prompt	string	Yes	Editing instruction
mask	file	No	Mask whose transparent areas are editable (DALL·E 2)
model	string	No	gpt-image-1 / dall-e-2
n / size / quality	-	No	Same as the generation endpoint

Legacy Endpoints

POST /v1/images/variations (image variations, DALL·E 2) and POST /v1/edits (legacy path) are proxied as well — parameters are the same as above. Use them only for older codebases.

tip

Models like Nano Banana (Gemini images) work through the OpenAI format too — just set model to the corresponding name; unrecognized parameters are passed through to the upstream.