AI flyer generator: make the art, then type the details

Generate flyer art at US Letter or A4 shape with empty panels, then type the headline, date, prices, and address yourself so you can check each one.

5 min readSume
All posts

To make a flyer with AI, generate the artwork at your page's shape with empty panels for the words, then type the headline, date, prices, and address in a layout tool, because an image model can misspell the small print it draws. On Sume, ChatGPT Image 2.5 takes a custom size such as 2528 × 3280 pixels, within 1% of a US Letter page's shape and about 297 pixels per inch on it, and today returns up to 4 designs per call.

Sume facts come from the Image API docs and the request schemas in the Sume API reference, read on 2026-09-28. Paper sizes and pixels per inch are plain arithmetic.

What size should an AI flyer be?

Generate at the shape of the paper you will print on, so nothing gets cropped, and divide pixels by inches to get pixels per inch. The table lists the largest ChatGPT Image 2.5 custom sizes within 1% of each paper's shape; the Letter and A4 rows are the same page sizes AI coloring page generator uses for printables.

A full Letter or A4 page lands just under 300 pixels per inch, while half letter and A5 land well above it. If your print shop asks for 300, upscale the art before you add the words with Image Upscale 1.0, which takes an upscale_factor from 1 to 4: 1.01 for Letter or 1.03 for A4. The docs don't say how a decimal factor is rounded, so check the pixel size of the file you get back. AI image for print explains the math and the upscale request.

Sizes computed from the custom-size rules in Image API, read 2026-09-28.
PaperSizeAsk ChatGPT Image 2.5 forPixels per inch
US Letter8.5 × 11 in2528 × 3280About 297
Half letter5.5 × 8.5 in2320 × 3568About 420
A4210 × 297 mm2432 × 3408About 291
A5148 × 210 mm2416 × 3424About 414

What should a flyer prompt ask for?

Describe the page as a layout with a picture in it, not as a finished flyer. Send it to POST /v1/images with openai/gpt-image-2.5 and the page size as image_size:

  • The layout: "a large photo area in the top half, a plain cream panel across the lower third, and a thin empty band at the bottom for contact details".
  • The subject and style: "fresh bread and pastries on a wooden counter, morning light, soft watercolor".
  • Your colors, by name, so the panels match your brand.
  • "No text, no letters, no logos", so the panels stay empty for your words.
  • n to compare up to 4 layouts per call. Each completed image is billed.
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": "Flyer artwork for a neighborhood bakery opening: fresh bread and pastries on a wooden counter in the top half, morning light, soft watercolor. A plain cream panel across the lower third and a thin empty band at the bottom. No text, no letters, no logos.",
    "image_size": { "width": 2528, "height": 3280 },
    "n": 4
  }'

Why not let the AI write the flyer text?

Because a flyer's details must be exact, and generated lettering can come out with a wrong character in the flat PNG, JPEG, or WebP you get back today, which has no text layer to correct, as AI image with text explains. Typed in a layout tool, a phone number, a price, or a street name can be checked, resized, and changed for next month's event without new art.

Add any QR code in the layout tool as well, made from your link with a QR code tool: a code drawn by an image model is only a picture of one.

Can I make a business flyer from my own photos?

Yes. Put a photo of your product, shop, or venue at a public HTTPS URL and send it in input_references, with a prompt that says how to use it: "the storefront from the photo in the top half, redrawn as a bright illustration". Keep image_size in the request: it wins over aspect_ratio, so the flyer keeps its page shape. The model redraws what it sees, so a logo or label in the photo can change. Place your real logo in the layout tool.

What does it cost, and what are the limits?

  • Each completed image is billed, and a failed or canceled one is not. ChatGPT Image 2.5 estimates each request's price from its size and quality, and usage.cost in the response is the USD amount billed. AI image generator API cost compares models.
  • High quality, this model's default, and a large n are two of the slow settings most likely to return 202 with a job instead of the images. Poll the job's status until it is terminal, then read its result.
  • Results are Sume-hosted, signed URLs, so download the files you keep.

Sources

Related posts

More in Use cases

All Use cases posts

Written by Sume