Reference

POST /v1/images/generations

Make pictures from a text prompt, in OpenAI’s format: the SDK’s images.generate works with only its base URL changed. Pictures come back in the response, and we keep no copy.

Checked 6 November 2026

Make a picture

import base64
import os
from openai import OpenAI

client = OpenAI(
    base_url="https://nokeep.ai/v1",
    api_key=os.environ["NOKEEP_API_KEY"],
)

result = client.images.generate(
    model="z-image-turbo",
    prompt="A lighthouse on a rocky coast at dusk",
    size="1024x1024",
)

with open("picture.webp", "wb") as f:
    f.write(base64.b64decode(result.data[0].b64_json))

Models

Model	Good for	Per image	Size	Longest prompt
chroma	An open model that reads long, detailed prompts.	$0.03	Up to 1280 pixels a side, returned at 1.5x	7,500 characters
z-image-turbo	The fastest here, for photographic pictures at volume.	$0.03	Up to 1280 pixels a side	7,500 characters
wai-illustrious	Anime and illustration.	$0.03	Up to 1280 pixels a side	1,500 characters
krea-2-turbo	Holds known characters and a house style, at 1K or 2K.	$0.08 (1K), $0.12 (2K)	1:1, 3:2, 16:9, 21:9, 9:16, 2:3, 3:4, 4:5	5,000 characters

GET /v1/models lists the same prices and limits.

Request

Field	Takes	What it does
`model`	Required	One of the model ids above.
`prompt`	Required	What to draw, sent to the model as written. Over the model’s limit is refused, never cut.
`size`	`WIDTHxHEIGHT` or `auto`	The picture’s size. The default is 1024x1024. See Sizes.
`aspect_ratio`	1:1, 3:2, 2:3, 4:3, 3:4, 4:5, 16:9, 9:16, 21:9	A shape, in place of `size`. It wins when both are sent.
`n`	1 to 4	How many pictures. You pay for those that come back.
`response_format`	`b64_json` or `url`	The default is `b64_json`. `url` returns a `data:` URL, not a link to a stored file.
`output_format`	`webp`, `png` or `jpeg`	The default is `webp`.
`resolution`	`1K` or `2K`	`krea-2-turbo` only, and priced by it.
`seed`	A whole number	The same seed and prompt give a similar picture. With `n`, each picture adds one to it.
`negative_prompt`	Text	What to leave out. It isn’t read by the hard-limit check, and not every model takes it into account.
`steps`	1 to 50	More steps take longer. Not every model uses it.
`cfg_scale`	Above 0, up to 20	How closely to follow the prompt. Not every model uses it.

Any other field, such as quality, style or user, is dropped and named in the x-nokeep-ignored header. A field that carries a picture, such as image, is refused with 400 unsupported_input: these models work from text only.

Sizes

A side can be at least 256 pixels, and the longer side no more than three times the shorter. Outside that is 400 unsupported_size.
A larger size is fitted inside 1280 pixels a side, keeping its shape. OpenAI’s 1536x1024 is drawn at 1280x848, for example.
chroma returns 1.5x the pixels they draw at, at no extra cost, so 1024x1024 comes back at 1536x1536.
krea-2-turbo draws in named shapes, so a size becomes the nearest one.
The response’s size or aspect_ratio always says what you got.

Response

JSON

{
  "created": 1793960000,
  "data": [
    {
      "b64_json": "UklGRk…"
    }
  ],
  "model": "z-image-turbo",
  "output_format": "webp",
  "size": "1024x1024",
  "usage": {
    "images": 1,
    "nokeep_credits": 30
  }
}

The x-nokeep-credits and x-nokeep-balance headers say what the call cost and what is left, as on chat. A picture is billed when it comes back. A failed or refused one costs nothing.

What the models will make

Adult content is in scope on the models the host marks uncensored. z-image-turbo isn’t one of them, and held back the most in our checks, so it can come back softer than asked.

Four hard limits stay: sexual content involving minors, sexual content of real people, mass-casualty harm and targeting a real person. A prompt that crosses one gets 422 hard_limit and costs nothing. The check reads the prompt; see Hard limits. A model’s host can also refuse a request or hold back a result: that is 422 content_refused, also free.

Limits and errors

A key can make 60 image requests a minute, inside its 120 requests overall. Past that is 429 rate_limited with Retry-After. The errors page lists every code.

To check your code without spending anything, send sk-nokeep-test as the key. It runs every check above and answers with a one-pixel picture.

curl

curl https://nokeep.ai/v1/images/generations \
  -H "Authorization: Bearer $NOKEEP_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "chroma",
  "prompt": "A lighthouse on a rocky coast at dusk",
  "n": 2,
  "aspect_ratio": "16:9",
  "output_format": "png",
  "seed": 42
}'

Get an API key Try the free test key: sk-nokeep-test