Create images · bInference Docs

POST

Send a prompt and an image model, get the images back as base64. Pick a model and its fields from GET /images/models: each model takes its own set, and a field it doesn't list is refused before anything is reserved.

The call reserves every image it asks for at the dearest host's price. Set resolution and quality to reserve less. See Images for costs, references and streaming.

Send your key as Authorization: Bearer binf_... or x-api-key.

Request body

The fields most calls use. Every other field of the OpenAI format is passed on as sent.

modelstringRequired
An image model id from GET /images/models, such as openai/gpt-image-2.
promptstringRequired
What to draw.
ninteger
How many images, 1 by default. Each model sets its own most, up to 10: see parameters.n in GET /images/models. The call reserves every image.
resolution"512" | "1K" | "2K" | "4K"
The size tier, for models that list it. Left out, the call reserves the largest tier the model offers.
aspect_ratiostring
Such as "1:1" or "16:9", from the values the model lists.
qualitystring
low, medium, high and up, for models that list it. Lower quality reserves and costs less. Left out, the call reserves the highest the model offers.
sizestring
A tier such as "2K", or pixels such as "2048x2048".
output_formatstring
png, jpeg, webp or svg, where the model offers a choice.
backgroundstring
transparent or opaque, for models that list it.
input_referencesobject[]
Images to edit or follow, as { "type": "image_url", "image_url": { "url": "https://..." } }, up to the model's most. Links are best: a request body is at most 4 MB.
streamboolean
Send partial images as they form, then the finished one, as server-sent events. One image a call, on models with streaming: true. Others answer the whole image as usual.
seedinteger
The same seed and prompt give the same image, on models that list it.
response_format"b64_json" | "url"
url stores the images and answers with a link to each, good for 7 days, instead of base64: a small answer however large the images. Not with stream.

Response

createdinteger
When it was made, in Unix seconds.
dataobject[]
The images, in order.
- b64_jsonstring
  The image, as base64. Left out with response_format: "url".
- urlstring
  The image's link, with response_format: "url".
- expires_atstring
  When the link stops working, with url.
- media_typestring
  Its type, such as image/png.
usageobject
Tokens and cost: what this call is charged, in dollars. Exactly what leaves the balance.

Response headers

x-binference-call-idstring
This call's id on your Calls page. Quote it when you contact us.
x-generation-idstring
The model provider's id for the answer.
retry-afterinteger
On a 429 or 503: whole seconds to wait before sending again.
retry-after-msinteger
The same wait in milliseconds. The OpenAI and Anthropic SDKs read it.

Request

curl https://binference.io/api/v1/images \  -H "Authorization: Bearer $BINF_API_KEY" \  -H "Content-Type: application/json" \  -d '{    "model": "openai/gpt-image-2",    "prompt": "A small red lighthouse on a rock at dusk, flat illustration",    "quality": "medium",    "aspect_ratio": "1:1"  }'

Response

{  "created": 1790809771,  "data": [    {      "b64_json": "iVBORw0KGgoAAAANSUhEUgAAB...",      "media_type": "image/png"    }  ],  "usage": {    "prompt_tokens": 15,    "completion_tokens": 1584,    "total_tokens": 1599,    "cost": 0.057114  }}

Streamed, with stream: true

data: {"type":"image_generation.partial_image","partial_image_index":0,"b64_json":"iVBORw0KGgo..."}data: {"type":"image_generation.completed","b64_json":"iVBORw0KGgo...","media_type":"image/png","created":1790809864,"usage":{"prompt_tokens":15,"completion_tokens":1584,"total_tokens":1599,"cost":0.057114}}data: [DONE]