OpenAI Generations

/v1/images/generations is the OpenAI Images API compatible synchronous image generation endpoint. The HTTP request completes after the image result is ready.

Interface

Item	Value
Method	POST
Path	/v1/images/generations
Content-Type	application/json
Authentication	Authorization: Bearer YOUR_API_KEY
Platform requirement	General API keys require an OpenAI platform group with image generation enabled. Task API keys can use this synchronous image endpoint.

API Key Behavior

Key type	Behavior
General API key / OpenAI group	Uses the synchronous OpenAI Images gateway with the existing behavior.
Task API key	Uses Task API models and pricing, waits synchronously for up to 120 seconds, then returns an OpenAI Images style response.

Task API keys support POST /v1/images/generations and POST /v1/images/edits, including /images/generations and /images/edits aliases without the /v1 prefix. They do not support /v1/responses or conversational endpoints. If the synchronous wait times out and the response includes task_id, use the Task API to query the final status.

Request Parameters

Parameter	Type	Required	Description
model	string	No	Image model. Defaults to gpt-image-2. General API keys must use an OpenAI Images compatible model; Task API keys may use any available Task API image model.
prompt	string	Yes	Image generation prompt.
size	string	No	Output size, such as 1024x1024.
n	integer	No	Number of images. Defaults to 1 and must be greater than 0.
response_format	string	No	Response format. General API keys commonly use b64_json; Task API keys support url and b64_json, defaulting to url.
stream	boolean	No	Whether to stream the response when supported by the selected model. Task API keys do not support stream=true on this endpoint.
quality	string	No	OpenAI Images API parameter. Availability depends on the selected model.
background	string	No	OpenAI Images API parameter. Availability depends on the selected model.
output_format	string	No	OpenAI Images API parameter. Availability depends on the selected model.
output_compression	integer	No	OpenAI Images API parameter. Availability depends on the selected model.
moderation	string	No	OpenAI Images API parameter. Availability depends on the selected model.
style	string	No	OpenAI Images API parameter. Availability depends on the selected model.

Request Example

curl https://api.llmapi.site/v1/images/generations \
  -H 'Authorization: Bearer YOUR_API_KEY' \
  -H 'Content-Type: application/json' \
  -d '{
    "model": "gpt-image-2",
    "prompt": "a product photo of a ceramic coffee cup on a white table",
    "size": "1024x1024",
    "n": 1,
    "response_format": "b64_json"
  }'

Response — 200 OK

{
  "created": 1779912000,
  "data": [
    {
      "b64_json": "..."
    }
  ]
}

Field	Description
created	Unix timestamp when the response was created.
data[].b64_json	Base64 image payload when response_format is b64_json.
data[].url	Image URL when URL responses are available.

With a Task API key, url returns an authenticated image asset URL when URL responses are available. Use response_format: "b64_json" when the client needs inline image bytes.

Errors

Status	Cause
400	Request body is invalid, the model is not an image model, or required parameters are missing.
401	API key is missing or invalid.
403	Balance, quota, or group permission check failed.
404	Current API key group cannot call the OpenAI Images API.
429	Rate limit or concurrency limit was reached.
500 / 502 / 503	Gateway or upstream provider failed.
504	The Task API key synchronous wait timed out. The response includes task_id when the final status can be queried later.

Task API key timeout example:

{
  "error": {
    "type": "TASK_SYNC_TIMEOUT",
    "message": "task did not complete before synchronous wait timeout",
    "task_id": "task_..."
  }
}

Boundaries

• This endpoint is only for OpenAI Images API compatible requests.
• Use Task API for long-running or production long-running jobs.
• A Task API key can use the OpenAI Images compatible generation and edit endpoints, but it cannot call text or conversation endpoints.