Formats · July 2026 · 6 min read

WebP vs PNG: Which Should You Use?

The short answer is that WebP is smaller, but the short answer is misleading. WebP is really two different formats sharing a name, and PNG still wins in a handful of specific, predictable situations. Here's how to tell which one you're in.

WebP is two codecs wearing one name

Most confusion about WebP starts here. Lossy WebP, released in 2010, is derived from the intra frames of the VP8 video codec; it compresses like a video keyframe and throws data away to do it. Lossless WebP, added in 2012, is a completely separate codec designed from scratch for lossless work, and it shares nothing with the lossy one except the file extension.

Most people only know the lossy one, so "WebP" has a reputation as a photo format that smudges fine detail. That reputation belongs to only one of the two codecs. The other is one of the best lossless image compressors in common use, and it is badly underused.

What PNG gets right

PNG dates to 1996 and is lossless by definition: every pixel comes out exactly as it went in. Under the hood it runs DEFLATE compression over per-row prediction filters, where each row is predicted from its neighbors and only the difference is compressed. It stores 8 or 16 bits per channel and has full alpha transparency.

Its biggest advantage, though, isn't technical. After roughly 25 years, PNG is supported everywhere: every browser, OS preview, editor, upload form, and printer portal ever written can open it. No other lossless format can say that, and it matters more often than compression ratios do.

The chroma subsampling problem

This is the most important thing to know before reaching for WebP: lossy WebP only supports 4:2:0 chroma subsampling. In plain terms, the format splits an image into brightness and color, keeps brightness at full resolution, and stores the color at a quarter of the resolution, one color sample for every 2x2 block of pixels. There is no quality setting, however high, that turns this off.

For photographs this is invisible. Your eyes are far more sensitive to brightness than to color, and photos rarely have color changing sharply from one pixel to the next, which is why video codecs (where the technique comes from) have used it for decades.

Point it at a sharp colored edge, though (red text on white, a UI screenshot, line art, a logo), and the quarter-resolution color smears across the edge. You get fringing and bleeding around every colored letter, and raising the quality slider to 100 does not fix it, because the data was never stored. If you've ever wondered why a screenshot saved as WebP looks subtly greasy around the text, this is why.

Lossless WebP: the underused one

The 2012 lossless codec has none of these problems. It is pixel-perfect, handles alpha, and typically produces files around 25% smaller than an optimized PNG of the same image. For graphics, logos, and screenshots delivered on the web, it is simply a better PNG.

Two caveats keep it from being a full replacement. It is 8-bit only, where PNG goes to 16 bits per channel, so it can't serve as an editing or print master. And while browser support is universal, support in the long tail of desktop software, enterprise systems, and image pipelines is noticeably weaker than PNG's.

Try it Convert & Compress Convert between WebP, PNG, and JPEG in your browser. Nothing is uploaded.

Where PNG is still the right choice

Where WebP wins

WebP wins at web delivery. Anywhere you control the display context (your own site, your own app), it is the better format to ship for both kinds of image: lossy WebP for photos, lossless WebP for graphics. Browser support has been universal since Safari 14 landed around 2020, so the old habit of keeping a PNG fallback is no longer necessary for browsers.

For photos, "WebP vs PNG" is usually the wrong comparison in the first place. PNG was never the right format for photographs, because lossless compression makes photo files balloon. The real photo-on-the-web contest is lossy WebP (typically 25–35% smaller than an equivalent-quality JPEG, with alpha support JPEG lacks) against JPEG and AVIF, which is the stronger modern option for photos. PNG only enters the conversation for graphics, screenshots, and masters.

The decision in one table

Use case Best choice Why
Photo for the web Lossy WebP (or AVIF) Far smaller; subsampling invisible in photos
Graphic, logo, or screenshot for the web Lossless WebP Pixel-perfect, ~25% smaller than PNG
Anything with unknown recipients PNG (JPEG for photos) Opens everywhere, no exceptions
Editing master PNG (16-bit) or TIFF WebP is 8-bit only
Favicon PNG / ICO The expected convention
Text-heavy screenshot you'd save as lossy WebP Use lossless WebP or PNG instead 4:2:0 subsampling fringes colored text
Frequently asked questions
01

Is WebP always smaller than PNG?

No, but lossless WebP usually is, typically around 25% smaller than a well-optimized PNG of the same image. Lossy WebP is far smaller still, but that is not a fair comparison, because it discards data and PNG by definition does not.

02

Why does my WebP screenshot look blurry around colored text?

Because lossy WebP always uses 4:2:0 chroma subsampling, which stores color at a quarter of the image's resolution, and no quality setting changes that. Sharp colored edges fringe and bleed as a result. Re-export as lossless WebP or PNG and the problem disappears.

03

Does converting PNG to WebP lose quality?

It depends on the mode. Lossless WebP is pixel-identical to the source PNG, so nothing is lost. Lossy WebP discards data by design, so quality does drop; that is where its much smaller files come from.

04

Can I use WebP for my website's favicon?

Browsers increasingly accept it, but PNG and ICO remain the safe convention, the formats every browser, crawler, and pinned-tab implementation expects. Our favicon generator produces the full conventional icon set from any image.

Keep reading Why AVIF Beats JPEG and WebP the next step beyond WebP Why Apple Uses HEIC Instead of JPG the mobile side of the story Favicon Generator a complete icon set from any image