AI-generated images often come out between 1,000 and 2,000 pixels wide, while a t-shirt needs about 4500 × 5400 pixels. There are two fixes: AI upscaling, which rebuilds detail for rich illustrations, and vectorizing, which turns the image into crisp shapes at any size, ideal for flat-color designs. Here is when to choose each, which tools to use, and the checks to run before printing.

Lire en français : Agrandir et vectoriser une image IA pour l'impression

Side-by-side comparison of a pixelated illustration and the same image upscaled sharp, magnifying glass in the center, technical style

Why AI images are too small

A 1,024-pixel image printed 12 inches wide gives about 85 dots per inch, far from the 300 recommended for sharp printing. The result: blurry edges and visible jaggies. Sizes per product are in our print file guide.

Some recent generators offer high resolution output directly (up to 2K or 4K depending on the tool and plan). That is a good start, but you often still need to upscale to reach 4500 × 5400 pixels.

Option 1: AI upscaling

An upscaler analyzes the image and invents the missing detail instead of stretching pixels. It suits detailed illustrations, textures, paintings and 3D renders.

ToolTypeGood to know
UpscaylFree, open source, desktopWorks offline, several models to choose from
Photoshop (Super Resolution)PaidDoubles the size while keeping detail well
Online upscalersFree or credit-basedHandy, but check what they do with your files

Tips: upscale in two passes rather than one (×2 then ×2), then scale down to the exact size. Check at 100%: an upscaler can create odd textures on flat areas.

Option 2: vectorizing

Vectorizing turns pixels into curves. The file (SVG, EPS, PDF) stays perfectly sharp at any size. It suits flat-color designs, thick outlines, logos, lettering and "sticker" or linocut styles.

  • Vectorizer.AI: online, very accurate on flat illustrations.
  • Illustrator (Image Trace): complete, with color count control.
  • Inkscape (Trace Bitmap): free, effective on black and white.
  • Vexels: its vectorizer sets the color count, removes the background and adds an outline.

Tips: limit the number of colors (3 to 8), remove small stray specks, then export a high resolution PNG if the platform does not accept vectors.

Vectorized sticker-style t-shirt design with paths and anchor points visible on a software screen

How to choose

  • Rich illustration, gradients, textures: AI upscaling.
  • Flat colors, outlines, logo, text: vectorizing.
  • Hybrid design: vectorize the graphic elements, upscale only the illustration, then assemble.

Plan for size in the prompt: ask for "vector", "flat colors" and "thick outlines" if you intend to vectorize. The result will be much cleaner.

Pre-print checklist

  1. Final size reached (for example 4500 × 5400 pixels for a t-shirt).
  2. 100% zoom: no blur, no stray texture, no halo.
  3. Truly transparent background (test it on a colored backdrop).
  4. RGB colors, with no unintended semi-transparent areas.
  5. Mockup on a light and a dark shirt.

Frequently asked questions

What is the best resolution to print an AI image on a t-shirt?

About 4500 × 5400 pixels as a transparent PNG, which is 300 dots per inch over a 15 × 18 inch area.

Should I upscale or vectorize?

Upscale detailed illustrations, and vectorize flat-color designs, logos and text.

Is there a free tool to upscale an AI image?

Yes. Upscayl is free, open source and runs on your computer, even offline. Inkscape is free for vectorizing.

In short

An AI image always needs preparing before print: upscaling for rich illustrations, vectorizing for flat colors. Plan for it in the prompt, check at 100%, and keep the vector or high resolution version as the master file for all your platforms.

Lire en français : Agrandir et vectoriser une image IA pour l'impression