AI coloring page generator: from a prompt or a photo

An AI coloring page generator draws black-and-white line art from a description or a photo: ask for clean black outlines on a plain white background, with no shading, gray, or color, in a portrait shape that fits your paper. On Sume, ChatGPT Image 2.5 takes a custom size such as 2528 × 3280 pixels, about 297 pixels per inch on a US Letter page; to start from a photo, send the photo as a reference with the same line-art instructions.

Sume facts come from the Image API docs and the request schema in the Sume API reference, read on 2026-09-27. The pixels-per-inch figures are plain arithmetic; AI image for print explains it for any print.

How do I write a coloring page prompt?

Describe the subject, then spell out the line-art style. Useful phrases:

• Black outlines only, on a plain white background.
• No shading, no gray tones, no color, no large filled black areas.
• Thick, closed lines and big open shapes for young children, or fine detailed patterns for adults.
• One subject centered on the page, with a white margin around it.

curl -X POST "https://api.sume.com/v1/images" \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "openai/gpt-image-2.5",
    "prompt": "Coloring page for kids: a friendly dinosaur in a jungle. Thick black outlines on a plain white background, closed shapes, no shading, no gray, no color",
    "image_size": { "width": 2528, "height": 3280 },
    "output_format": "png"
  }'

How do I turn a photo into a coloring page?

Send the photo in input_references as a public HTTPS URL, with the same style instructions in the prompt. The Image API docs use the same pattern to turn a photo into a watercolor painting. ChatGPT Image 2.5 takes up to 16 references, and keeping image_size in the request makes the page portrait even when the photo is landscape.

The model draws a new image from the photo, so faces and small details can change. Check the likeness before you print.

{
  "model": "openai/gpt-image-2.5",
  "prompt": "Turn this photo into a coloring page: black outlines on white, no shading, no color",
  "input_references": [
    { "type": "image_url", "image_url": { "url": "https://example.com/dog.jpg" } }
  ],
  "image_size": { "width": 2528, "height": 3280 }
}

What size should a coloring page be?

A portrait page in your paper's shape, with enough pixels for the print: divide the pixels by the inches to get pixels per inch. If you need more pixels than a custom size allows, send the page's public HTTPS URL to Image Upscale 1.0, which enlarges it by an upscale_factor from 1 to 4.

A ChatGPT Image 2.5 custom size needs both edges in multiples of 16, at most 3840 on the long edge, and 655,360 to 8,294,400 pixels, so these are the largest sizes within 1% of each paper's shape:

Can I make a whole coloring book?

Yes, with one prompt per page. Each call can return up to 4 versions of that page with n, the model's ceiling today, so you can keep the cleanest; each image is billed. Its price is estimated from size and quality, and usage.cost in the response is the USD amount billed. Bulk image generation covers sending many prompts from a script.

High quality, the default on this model, and a large n are the slow configurations most likely to return 202 with a job to poll. Results are Sume-hosted, signed URLs in data[].url, so save each page you keep.

What should I check before printing?

Look at every page at full size:

• Lines are closed, so each area can be colored on its own.
• No gray shading or solid fills slipped in.
• Any words are spelled right; generated lettering can be misspelled.
• The page is portrait and matches your paper, so nothing is cut off when it prints.