![Why AI Images Look Blurry When Printed — and How to Fix It [Resolution & CMYK]](/blog/thumbnails/ai-print-blur.webp)
It looked crisp on screen, but came out grainy on paper
Printed an image you made in Midjourney or ChatGPT and found it strangely blurry? We hear this a lot.
It looks perfectly fine on a phone or computer, but turns grainy once it’s on actual paper. Plenty of people have planned to use one as the main visual on a flyer, only to be disappointed when the print arrived.
The cause is clear, and so is the fix.
”Resolution” is the root problem
Image resolution is how many dots are packed into one inch, measured in dpi (dots per inch).
For a phone or computer screen, 72–96 dpi is plenty to look sharp. Print, however, needs 350 dpi. That’s roughly 4–5 times as many dots packed into the same area.
Both Midjourney and ChatGPT output images at screen resolution by default. That’s why they look great on screen but go grainy in print.
The pixel-count trap
“But my image is huge,” many people think. Pixel count and resolution are two different things, though.
Midjourney’s default output is 1024×1024 px, but print that at A4 (210×297 mm) and the resolution is only about 123 dpi.
| Image size | Resolution at A4 | Print quality |
|---|---|---|
| 1024×1024 px | ~123 dpi | Grainy |
| 2048×2048 px | ~247 dpi | Still not enough |
| 2912×4096 px (Midjourney max) | ~350 dpi | Passing grade |
The first thing to do is use upscaling and export at the largest size before you use the image.
Calculate the pixel count you need yourself
“For this print size, how many pixels do I actually need?” You can work that out with a simple formula.
Required pixels = print length (mm) ÷ 25.4 × 350
Here, 25.4 converts inches to mm, and 350 is the dpi print needs. For example, A4’s 297 mm long edge gives you 297 ÷ 25.4 × 350 ≈ 4093 px as a target. Here are the required pixel counts for common print sizes.
| Print size | Required pixels (350 dpi) |
|---|---|
| Business card (91×55 mm) | ~1254×758 px |
| Postcard (100×148 mm) | ~1378×2039 px |
| A5 (148×210 mm) | ~2039×2894 px |
| A4 (210×297 mm) | ~2894×4093 px |
| A3 (297×420 mm) | ~4093×5787 px |
| B2 poster (515×728 mm) | ~7098×10031 px |
If your generated image has fewer pixels than the figure shown, it’s likely to look blurry in print. Conversely, if you’re printing small, you can get away with somewhat fewer pixels — a judgment call you can now make for yourself.
How far can upscaling actually “recover”?
Here’s an easy misconception: “I can just upscale anything to 350 dpi.” AI upscalers don’t add information that isn’t there — they guess at the missing detail and fill it in. So:
- About 2× enlargement: Usually clean, and can often be lifted to print quality without breaking down.
- 4× or more: Thin lines and fine patterns can flatten out or start looking unnatural.
In other words, “generate big from the start” is the real answer, and upscaling is only a backup. Force a too-small image to stretch and the blur may vanish, but you’ll trade it for a flat, “coloring-book” look.
It’s not just resolution — the color changes too
It’s easy to overlook while you’re focused on the blur, but the color shifts as well.
AI images are generated in RGB (the three primary colors of light). Print builds color in CMYK (four inks). The two have different color ranges, so when you hand RGB straight to a print shop, colors get dulled or shifted during automatic conversion.
The resolution problem and the color problem happening at the same time is a very common situation with AI images in print. For how the color shifts and why CMYK is necessary, see What Is CMYK? The Difference from RGB and Why Print Submission Needs It.
How to handle it without Photoshop
The most reliable route is to raise resolution and convert to CMYK in Photoshop — but that means a monthly subscription and a learning curve. Signing up just for a single flyer is a high bar.
The free browser tool DecoTecoPop handles CMYK conversion and automatic 3 mm bleed just by uploading an image — and PDF/X-4 export is free while we’re in beta. No sign-up, no install, and it works from a phone too.
As for resolution, the premise is to generate at the largest size you can before uploading. DecoTecoPop handles the resolution of the image it receives faithfully, so if the source is small the print quality will be modest. The key is to export at maximum size first, then convert.
Things to keep in mind from the moment you generate
Generate at the largest size possible
In Midjourney, upscale (U1–U4) before exporting. With ChatGPT, asking it to “output in high resolution” tends to produce something larger.
Decide the aspect ratio for your print size first
Generate at an aspect ratio close to your print size — --ar 3:4 for A4 portrait, --ar 4:3 for A4 landscape — so you avoid cropping later and losing resolution.
Do the CMYK conversion yourself before submitting
Leaving it to the print shop means you can’t check the color. It’s safer to convert to CMYK with some tool beforehand and see the converted color with your own eyes before submitting.
FAQ
Q. Which phones and browsers does it work on?
CMYK conversion, bleed, and PDF creation for your image work in recent phone browsers (Safari or Chrome). If you also want to use the text-decoration editor, you’ll need a device that supports WebGPU (iOS 26 or later in Safari on iPhone, or Chrome on Android).
Q. Can the resolution of an AI image be increased?
Using an AI upscaler (such as Real-ESRGAN) can raise it to a degree. It isn’t perfect, but it enlarges a 1024 px image 2× or 4× while reconstructing detail, so it prints better than the original. We plan to add this feature to DecoTecoPop as well.
Q. Is it OK to submit to a print shop in RGB?
Some print shops will accept it and convert to CMYK on their end. But you can’t check the converted color, so there’s a risk it won’t come out as intended. If color matters, it’s safer to convert yourself before submitting.
Q. dpi and ppi — the resolution units keep getting mixed up. Are they different?
They’re used to mean almost the same thing. Strictly speaking, ppi (pixels per inch) is the density on the image-data side, and dpi (dots per inch) is the density on the printing-machine side. For submission, though, you can treat “a 350 dpi image = a 350 ppi image” with no problem. What matters is whether you have enough pixels for the print size.
Q. If I submit while the image is still small, can the print shop fix it?
As a rule, no — a print shop won’t make up the resolution itself. A small image stays small and gets stretched, printing grainy. So “generate large” and “upscale if it’s short” are jobs to do on your side before submitting.
Q. Which looks grainier — photos or illustrations?
Illustrations and logos, with their crisp lines and flat color areas, tend to show the jaggies of insufficient resolution more clearly. Photos have lots of gradation to begin with, so even at slightly lower resolution they can still look “sort of okay.” The more an AI image contains logos or text, the more attention you should pay to resolution.
The prep needed to make printed material from an AI image is four steps: “generate large → convert to CMYK → add bleed → export PDF.” Keep that flow in mind from the start and you’ll cut down on post-print disappointment considerably.
→ Build your print data with DecoTecoPop (PNG / JPEG / WEBP — up to 60 MB — no sign-up)