Why Is My PNG So Large? (And How to Fix It)

Nothing went wrong — the PNG was doing exactly what it was built to do, and that was the problem: it never throws a single pixel away, and that honesty costs bytes. A lot of them, on the wrong kind of image.

Treating "my PNG is huge" as one problem is the first mistake. A logo, a screenshot, and a photo are three different problems wearing the same file extension — and each one has an obvious, usually free fix. Most guides skip the question of what's inside the file entirely.

Why a PNG can't help being big

PNG compresses with DEFLATE — the same algorithm inside ZIP files — after a filtering step that rewrites each row of pixels as differences from its neighbours. Flat areas turn into long runs of near-zero values, and those compress beautifully. A photograph balloons because it has almost no repeating patterns for DEFLATE to grab onto: millions of slightly different pixels have to be recorded more or less one by one.

There's also no quality slider to rescue you, and this trips up everyone who comes from JPG. A PNG compression level from 0 to 9 only changes how hard the encoder works at finding patterns — it never changes a pixel. Most apps export at a fast middle setting, so your PNG is usually not even packed as tightly as it could be.

Pixel count is the one nobody checks first

It is also the most overlooked: a 1920×1080 image holds about 2 million pixels; at 32-bit colour with transparency that's roughly 8 MB of raw data before compression even starts. A Retina screenshot is worse: the capture is 2× or 3× the layout size, so you're storing pixels nobody will ever see. Before touching any compressor, check the dimensions — a 4000-pixel image displayed in an 800-pixel column carries five times the pixels it needs, and no amount of clever encoding fixes that.

Your logo doesn't need 16.7 million colours

Most PNGs are saved as 24-bit (16.7 million colours) or 32-bit (the same plus alpha). A logo with twelve colours does not need 16.7 million of them. Dropping to an indexed 256-colour palette — PNG-8 — routinely cuts flat graphics by 50–70% with no visible change. The fewer colours the image genuinely needs, the smaller the file.

The trap runs the other way: a photo or smooth gradient forced into 256 colours turns banded and ugly. And watch for 16-bit exports — they double the data per channel for detail nobody can see on a normal screen.

The twist: sometimes PNG is the small one

For flat graphics, PNG is often smaller than JPG — contrary to most articles that call PNG always bigger. One measured test put a UI screenshot at 164 KB as PNG and 132 KB as JPG — barely smaller, with smudged text — while PNG-8 came in at 58 KB, razor sharp. The gap explodes only on photos, where the same image can be 5–10× larger as PNG.

So the first question is never "how do I shrink this PNG?" It's "is this image even a PNG job?" Photos go to JPG or WebP. Drawings go to PNG. Get that backwards and you're fighting the format instead of using it.

Transparency costs you — except when it doesn't

The alpha channel is a fourth channel of data for every pixel, and it typically adds 15–30% to the file. If you don't need transparency, dropping it is free size. But there's an exception worth knowing: fully transparent pixels compress extremely well, so a logo on a transparent background can come out smaller than the same logo flattened onto white — because the editor fills the empty area with solid colour data. Transparency isn't automatically expensive. Unnecessary transparency is.

The last quiet culprit is metadata and interlacing. PNGs can carry ICC colour profiles, timestamps, text chunks, and EXIF data — usually 1–50 KB, occasionally several hundred KB straight out of a photo editor. Stripping them costs nothing visible. Interlaced PNGs, the ones that fade in gradually as they load, are also slightly larger, and on a modern connection the progressive preview isn't worth paying for.

The fastest way to make a giant PNG: convert a JPG into one

This is the single most common way people create giant PNGs. A 1 MB JPG photo converted to PNG can easily become 6 MB, because PNG now has to store every pixel of that photo losslessly — including every compression artefact the JPG already baked in. Converting back restores nothing; it just wraps the same imperfect pixels in a much bigger file.

If the file started life as a JPG, keep it as a JPG — or move to WebP. The only good reason to convert JPG to PNG is before editing, so later saves stop adding damage.

Five fixes, in the order worth trying them

1. Lossless re-encode — do this first, always. Run the PNG through a proper optimizer: oxipng or OptiPNG on the command line, or any decent online compressor. Nothing changes visually; the encoder just packs the same pixels tighter and strips junk chunks. Expect 10–30% off, more on screenshots and logos, less on photos. There's no downside: oxipng -o 4 --strip all input.png.

2. Cut the palette. If the image is a logo, icon, diagram, or flat UI graphic, quantize it to 256 colours with pngquant (pngquant --quality=65-80 input.png) or TinyPNG. This typically saves 50–70% while keeping transparency intact. Check gradients at full zoom before shipping — banding is the one visible risk, and dithering usually hides it.

3. Resize to the size it'll actually be shown at. For a Retina screenshot, this single step often beats every other trick. Resize to display width first, then compress.

4. Strip metadata and drop interlacing. One command handles the first: magick input.png -strip output.png. Small win on big files, meaningful on small icons where every kilobyte counts.

5. Change format when PNG isn't required. This is where the biggest numbers live. Google's measurements put lossless WebP about 26% smaller than the equivalent PNG with identical pixels, and lossy WebP at quality 80–85 shrinks a photographic PNG by 60–80% while keeping transparency. A photo with no transparency belongs in JPG at quality 80–85 — typically 70–90% smaller than the PNG. Our PNG to WebP guide covers which mode fits which image. Measured sizes for the same image in every format are in the WebP vs PNG vs JPG comparison.

Sometimes the big file is the right file

Don't optimize everything. Keep PNG when the image has text, sharp lines, or flat colour that JPG would smudge with halos; when it needs a transparent background and the destination doesn't accept WebP (some email clients and older tools still don't); and when it's a master copy you'll edit again — every JPG save adds a little damage, PNG saves add none. A 29 KB logo doesn't need optimizing. A 9 MB photo saved as PNG does.

Diagnose it in 30 seconds before you fix anything

Right-click the file and check its dimensions and bit depth. A 4000-pixel-wide file destined for an 800-pixel slot is a resize job. A 32-bit file with no transparency is a palette or alpha job. A 16-bit file from a photo editor is almost certainly an export-settings job — re-export at 8-bit and watch it shrink. And if the file is a photo wearing a .png extension, no amount of PNG optimization will beat converting it: that's a format job, not a compression job. If you need it converted, our Image Converter runs entirely in your browser, so screenshots with private content never leave your machine.

The rule of thumb that covers almost everything: photos go to JPG or WebP, drawings go to PNG, and anything you control on the web goes to WebP first. The format should match the content — not the app that happened to save it.