An NSFW image generation API: uncensored image models on one OpenAI compatible endpoint

Most image APIs refuse anything adult. OpenAI, Google and the big hosted APIs block nudity outright, and many "uncensored" sites only work in their own web app, with no API. OpenUncensored serves 17 uncensored image models, including the explicit Lustify, PornMaster and Pony checkpoints and the most popular anime models, through one OpenAI compatible endpoint. If your code already calls client.images.generate, change the base URL and it works.

The models

Model	id	Best for	Notes
Z-Image Turbo	z-image-turbo	Photoreal	Fast, cheap, follows plain English prompts. The best default.
Lustify v8	lustify-v8	Explicit photo	SDXL fine-tune built for explicit adult photography. The most permissive photo model.
Lustify v7	lustify-v7	Explicit photo	The previous Lustify release. Some prefer its skin tones.
Lustify SDXL	lustify-sdxl	Explicit photo	The original Lustify checkpoint.
Chroma	chroma	General	Uncensored FLUX.1-schnell derivative. Strongest at following long, detailed prompts.
Qwen Image	qwen-image	General	Best at rendering text inside the image (signs, captions, covers).
Qwen Image 3 Uncensored	qwen-image-3-uncensored	Photoreal	The most detailed output, at 1K or 2K. The priciest model here.
HiDream O1 Dev	hidream-o1-dev	Photoreal	Fast 2K model that allows adult content.
Juggernaut XL	juggernaut-xl	Photoreal	The most downloaded SDXL photoreal checkpoint. Great lighting and people.
CyberRealistic XL	cyberrealistic-xl	Photoreal	Photoreal SDXL, especially strong on people and skin.
PornMaster	pornmaster	Explicit photo	SDXL checkpoint for explicit adult photography.
Pony Diffusion V6 XL	pony-diffusion-v6	Stylised / Pony	The original Pony model for stylised and adult art. Needs score tags.
CyberRealistic Pony	cyberrealistic-pony	Stylised / Pony	Photoreal Pony fine-tune. Pony prompt tags with realistic output.
Pony Realism	pony-realism	Stylised / Pony	Realistic Pony fine-tune for adult content. Needs score tags.
WAI Illustrious	wai-illustrious	Anime	The most popular Illustrious anime checkpoint. Start with "masterpiece, best quality".
Hassaku XL	hassaku-xl	Anime	Illustrious based anime checkpoint for adult art.
NoobAI XL	noobai-xl	Anime	Illustrious derived base trained on Danbooru and e621 tags. Prompt with booru tags.

Models marked as Civitai checkpoints run on Civitai's own generator, and their creators get a share of every image.

Every model is routed to its cheapest provider, with automatic fallback to the next one if a provider is down.

Call it

curl https://openuncensored.com/api/v1/images/generations \
  -H "Authorization: Bearer $OU_API_KEY" -H "Content-Type: application/json" \
  -d '{"model": "z-image-turbo", "prompt": "a woman in a red dress on a rooftop at night, neon city behind her, cinematic", "size": "1024x1024"}'

The response is the standard OpenAI shape: data[0].b64_json (or data[0].url for some providers) plus an ou object with the provider and the cost of the request. With the OpenAI Python SDK:

import base64, urllib.request
from openai import OpenAI

client = OpenAI(base_url="https://openuncensored.com/api/v1", api_key="ou-...")

result = client.images.generate(
    model="lustify-v8",
    prompt="boudoir photo of a woman in her thirties, soft window light, 85mm, film grain",
    size="1024x1536",
)
image = result.data[0]
data = base64.b64decode(image.b64_json) if image.b64_json else urllib.request.urlopen(image.url).read()
open("boudoir.png", "wb").write(data)

Parameters

• model: an id from the table, or GET https://openuncensored.com/api/v1/images/models for the live list with prices.
• prompt: up to 8000 characters.
• size: 1024x1024 (default), 1024x1536, 1536x1024, 1024x1792, 1792x1024 or 512x512.
• n: 1 to 4 images. Each one is billed.
• negative_prompt: what to keep out of the image, for the Civitai checkpoints, for example "blurry, extra fingers, watermark".

Nothing is stored on our side: the image comes back in the response and is gone from our servers.

Which one to pick

• Photoreal, safe default: z-image-turbo. Fast and cheap, and good at plain sentences.
• Explicit photography: lustify-v8, or pornmaster and cyberrealistic-xl for a different look.
• Anime and hentai style: wai-illustrious or noobai-xl with booru style tags.
• Stylised and furry art: pony-diffusion-v6 with Pony score tags.
• Highest detail: qwen-image-3-uncensored at 1536x1024 or larger, which it renders at 2K.

The prompting guide compares them in more detail, with the prompt format each one expects.

No code? Use the desktop app

The OpenUncensored desktop app has the same image models built in as a tool. Ask its uncensored assistant for a picture in plain words and it writes the prompt, picks a model and saves the images to a folder on your Mac.

What is allowed

Adult content between adults is allowed on every model listed here. Three things are not, and the providers refuse them: anything involving minors or anyone who looks underage, real people (celebrities, people you know, lookalikes) without their consent, and non consensual imagery. If you publish what you make, label it as AI generated and follow the rules of the site you post on.

Get a key

No account and no sign up. Buy prepaid credits at openuncensored.com/credits and you get an ou-... key. The same key works for every text and image model, in the API, the playground and the desktop app. Each image is billed from your credits, typically one to five cents; the models page shows the live price of each one.