Image generation API

Updated 9.11.2026

Image generation is compatible with the OpenAI Images protocol. Only image models (priced per image in the model catalog) can be used; the chat endpoints do not accept image models.

Generate images

POST https://api.admesh.ai/v1/images/generations

curl https://api.admesh.ai/v1/images/generations \
  -H "Authorization: Bearer $ADMESH_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "<model-id>",
    "prompt": "A watercolor lighthouse on a cliff at dawn",
    "n": 1
  }'

Returns a standard OpenAI Images response; each item in data is one image (url or b64_json, depending on the model).

Request fields

Field	Description
`model`	Required, image model ID
`prompt`	Required, text prompt
`n`	Number of images, positive integer, default 1, at most 10 per request
`size`, etc.	Other OpenAI Images parameters are passed through where the model supports them

stream is not supported.

Billing

On submit, "images × price per image" is put on hold; with insufficient balance you get 402 insufficient_balance.
On completion you are billed for the images actually returned (never more than n) and the rest of the hold is released; tokens are not billed.
Failed generations are not billed.