Skip to main content
Blog

Pixels vs DPI: What Image Resizing Actually Changes

Somewhere in your inbox right now there is a client asking for a logo "as a 300 DPI PNG". They mean it as a quality setting. It is not one. Understanding what DPI actually is — and what it is not — saves you from blurry prints, 20 MB email attachments, and pointless debates about pixels.

The short version: pixels are the data, DPI is a suggestion. A 1200×600 image contains 720,000 pixels no matter what number lives in its metadata. The DPI tag only tells a printer or a page-layout program how large to spread those pixels on paper.

Pixels: the only thing your image is made of

Every raster image is a grid of pixels. Total width and height in pixels is the only measurement that describes the image itself. Change it and you either throw data away (downscale) or invent it (upscale, where the software guesses new pixels — usually badly).

  • Downscaling from 4000px to 1200px: fine. You discard detail the viewer could not see anyway at that size.
  • Upscaling from 400px to 1200px: the editor interpolates. Sharp edges soften, text gets fuzzy, JPEG artifacts inflate. No algorithm recovers data that was never captured.

Rule of thumb for the web: prepare images at roughly 2× the CSS size (a 400px-wide avatar slot gets an 800px file) so retina screens have real data to work with, then compress.

DPI: a printing instruction, not a property

DPI (more precisely PPI when we talk about images) says: "render these N pixels across one inch of paper." The same 1200×600 image tagged at 120 DPI prints at 10×5 inches; tagged at 300 DPI it prints at 4×2 inches. Nothing about the file changes except the tag.

Where the number 300 comes from: offset presses and typical office lasers can deposit about that many dots per inch, and a print held at reading distance needs that density to look continuous instead of dotted. It is a constraint of print, not of your file.

  • 300 DPI at 4×2 inches: requires exactly 1200×600 pixels.
  • The same pixels printed at 10×5 inches: effectively 120 DPI. Acceptable for a poster viewed across the room, unacceptable for a business card.

So "make it 300 DPI without resizing" usually translates to: "the file is too small for the sheet size." The fix is more pixels, not a metadata edit.

Resampling vs re-tagging

Two different operations that people constantly conflate:

  1. Re-tagging: change the DPI field from 96 to 300. File size stays nearly identical, pixel data untouched, the computed print size shrinks 3×. Pointless for web images that never get printed — but harmless, which is why so many stock libraries do it for customer reassurance.
  2. Resampling: actually computing a new pixel grid. This is what resizing does, and it is what changes file size and sharpness.

When a colleague sends you a 72 DPI image that looks "wrong" in Word or InDesign, the layout app is just reading the tag and computing a huge print size. Re-tag it instead of resampling upward and losing quality.

From practice: the 6000px photo problem

A modern phone photo is around 4000×3000 (12 MP) or more. Passing it to a CMS as-is means 4-6 MB, a slow page, and wasted mobile data. Performance audits flag images delivered substantially larger than their rendered size, and "properly size images" is one of the most common mobile issues. Downscale to the needed slot plus retina headroom, then compress: that same photo typically lands at 100-250 KB with no visible difference at phone viewing size.

This is exactly the pair of operations that image resizer and image optimizer handle when you work without a desktop editor: resize to the target CSS size ×2, then strip metadata and compress. Crop first if the aspect ratio needs fixing — run the image cropper before resizing, so you never ship dead margins inside the file.

From practice: the print shop asks for 300 DPI

A business card is 3.5×2 inches; with 0.125-inch bleed on each side that is 3.625×2.125 inches, which at 300 DPI requires about 1088×638 pixels per side. If the source logo is a 400px JPEG, no DPI edit fixes it. You either redraw it as vector (the right answer) or print it smaller than planned. The honest check: pixels ÷ 300 = the maximum clean-print width in inches. Run that division before promising a poster printed from a 900px avatar.

When inches, pixels and DPI get mixed on a non-square asset and you need the target pixel size fast, image dimension & DPI converter does that arithmetic directly, so you can confirm feasibility in seconds instead of dividing in your head at the print-shop counter.

The "export at 300 DPI for web" trap

Designers still get handed the request "a 300 DPI PNG for the website". Browsers ignore the DPI field completely; the CSS pixel size rules. Exporting "at 300 DPI" for web use simply means exporting at a huge pixel size — the worst option in practice: a heavy file that the browser then downscales in-page with its fast-but-basic algorithm. For screens, deliver at logical size ×1 and ×2, wire it into srcset, and let the browser choose.

A practical checklist

  1. Target is the web: work in pixels (CSS size ×1 and ×2), ignore DPI entirely, compress hard, use srcset.
  2. Target is print: required_pixels = inches × 300 per side. Not enough pixels? Redraw as vector or shrink the print size. Never upscale to fake it.
  3. A layout app complains about DPI: re-tag the field only; do not resample.
  4. Email attachments: cap the long side at 1000-1600px, which is plenty for reading on screen. A 6000px photo drops under 300 KB.

One check to close with: open any image's properties and read both the pixel size and the DPI field. If the pixels cannot cover the intended print size at 300, the tag value is furniture. Resize the pixel grid to fit the actual use, compress, ship.