How to enlarge AI art for print without it looking blurry

I use generated pictures as a starting point for some classroom decor, the way I might use a sketch. They are not finished print files. A picture that looks sharp on a phone can be a few thousand pixels on the long side. A 24×36 inch print at 300 dots in each inch needs 7200 × 10800 pixels. The gap between those two facts is the whole problem.

You can make the grid bigger. You cannot honestly say every new dot is detail that was already there. Some of it is a guess. If you know that before you list the poster, you can check the guess. If you skip the check, the buyer finds it on paper.

What "bigger" can mean

There are two different moves, and I keep them separate in my head.

The first move is a real enlargement step. A model looks at the small picture and proposes a larger one, often four times the width and four times the height. A 1000 × 1500 picture becomes 4000 × 6000. That is new pixels, chosen by the model. Simple shapes and soft painted areas often look better than a plain stretch. Faces can change in small ways. Text is where I expect trouble.

The second move is ordinary stretching after that. If 4000 × 6000 is still short of the sheet you want, a design app can scale the result again until the inch size matches. Those extra pixels are blending, not a second act of imagination. They will not repair a letter that the first step already got wrong. When I describe a file, I do not call that second scale "more AI." It is a resize.

A preview that is 1122 pixels wide and 1402 pixels tall is a useful example from my own tests. Four times that grid is 4488 × 5608. Divided by 300, that is about 15 by 18.7 inches at the sharp target. It is not yet a 24×36. Getting to 7200 × 10800 means another scale, or a decision to print at fewer than 300 dots per inch. I want to know which one I chose.

Lettering needs rebuilding

Spanish classroom art is full of words. A title, a plural, an ñ, an accent that changes the meaning. A model that barely has a mark where the accent sat will invent a mark. Sometimes the invention is close. Sometimes it is a different letter. I do not sell that as a finished poster.

My rule is simple. If a word must be read, I type it again after the enlargement. I put the enlarged picture on a locked layer. I set the type in a size I can measure against the chart. For a poster that will be 18×24 at 300 dots per inch, I am working toward 5400 × 7200, and the title has to look clean at that size, not only at the size of my laptop screen.

I read every line. I check the upside-down feeling of a missing accent by saying the word. "Sí" and "si" are not a style choice. Then I export the type as real vectors or as text rendered at the final pixel size. I do not run that new type back through another enlargement.

Illustrations without words are a kinder case. A big animal shape, a flat background, a simple border. I still compare a close crop of the original with the same crop of the enlargement. I look for extra texture that looks like crumpled paper, for edges that grew a halo, and for eyes that moved. If I would not hang it in my own classroom, I do not list it.

A check before anyone pays

I do three checks, in this order.

First, the pixel count. I divide by 300 and write the inches down. I only advertise sizes the file can support, or I clearly accept a lower dot count for a size people will see from farther away.

Second, a close look at the risky spot. One image pixel should map to one screen pixel while I inspect it. Fitting the whole poster on the screen hides the problem, because the screen shrinks it again.

Third, ink, when I can. A letter-size page printed at home, showing a piece of the poster at the real print size, tells me more than another zoom on a glowing display. Paper and a cheap inkjet are honest. If the small words break up there, they will not heal at the copy shop.

I also keep the original small file. If the enlargement goes waxy, I can try again or abandon it. I do not delete the only clean version.

What I promise, and what I do not

I can promise a buyer the pixel size, the file format, and which frame ratios are included. I can promise I looked at the lettering. I cannot promise that every enlargement recovered detail that was never in the small picture. I cannot promise a print shop, a paper type, or a frame I have not seen. The listing says what the file is. The close check is how I decide to sell it.

If you want a second pair of eyes on the numbers, the size check here uses pixels divided by 300 and does not treat a DPI label as magic. If you enlarge, look at the close-up of your own art before you download the full-size file. And if the poster has words, plan to rebuild them. That one extra step is the difference between a soft guess and a poster I would put on my wall.