AllConvert

Image formats explained: JPG, PNG, WEBP, HEIC and SVG

Every image format is a bet about what matters more: file size or fidelity. Once you understand which bet each format makes, choosing between them stops being guesswork. This guide covers the five formats you will actually meet in daily life, what each is good at, and — because this is where most people go wrong — a plain decision list at the end.

Lossy vs lossless: what compression actually does

There are two families of image compression, and they work in genuinely different ways.

Lossy compression, the JPG approach, starts from a fact about human eyes: we notice brightness far more than we notice fine color detail. A JPG encoder cuts the image into 8×8 pixel blocks and runs each block through a mathematical transform (the DCT, if you want the name) that describes the block as a mix of smooth gradients and fine wiggles. It then throws away most of the fine wiggles, because you would not have seen them anyway. The result is a much smaller file — and pixels that are no longer exactly what the camera recorded. That data is gone permanently. Saving the file again at higher quality does not bring it back.

Lossless compression, the PNG approach, is closer to how a ZIP file works. PNG first predicts each pixel from its neighbours (a row of sky-blue pixels is very predictable), then compresses the leftover differences with DEFLATE — the same algorithm inside ZIP. Every pixel survives exactly. The catch: this only saves space when the image has repetition to exploit. Flat areas, repeated colors and hard edges compress beautifully. The pixel-level noise in a photograph looks random to DEFLATE, so photos saved as PNG come out enormous.

JPG: the photo workhorse

JPG has been the default photo format since the mid-1990s, and for photographs it is still hard to argue with. Photos are full of soft gradients and subtle texture — exactly the content lossy compression handles gracefully. A 12-megapixel photo that would be 30 MB uncompressed lands around 3–5 MB as a high-quality JPG, and nobody can tell the difference on a screen.

Its weaknesses are just as specific. JPG has no transparency, so it cannot do cut-out logos or images that sit on a colored background. And it hates hard edges: text, line art and UI screenshots grow smudgy halos ("ringing") around every sharp boundary, because sharp boundaries are exactly the fine detail the encoder discards. If you have ever seen a blurry meme with crusty edges, that is a screenshot that lived its life as a JPG.

Use JPG for photographs going anywhere ordinary: email, web pages, documents, marketplaces. Avoid it for anything with text or sharp graphics.

PNG: exact pixels, transparency included

PNG is the opposite bet. Everything is preserved exactly, and it supports a full alpha channel, so a logo can have genuinely transparent corners rather than a fake white box. For screenshots, diagrams, charts and logos, PNG is the right answer almost every time — the text stays razor sharp and the file is often smaller than a JPG of the same screenshot, because flat interface colors compress so well losslessly.

The classic mistake is using PNG for photos. It works, in the sense that the file opens, but a photo that would be 2 MB as a JPG can be 15 MB as a PNG with zero visible benefit. If you have inherited photos stuck in PNG, running them through a converter in the other direction — or better, straight through our image compressor — usually shrinks them 5–10×.

WEBP: the newer all-rounder

WEBP, from Google, does both jobs: it has a lossy mode that typically beats JPG by 25–35% at the same visual quality, and a lossless mode that usually beats PNG. It supports transparency in both modes and even animation. For images published on the web, it is quietly the best default in 2026 — every current browser has supported it for years.

The drawbacks are about the world outside browsers. Some older desktop software, corporate tools and photo-print services still refuse WEBP files, and a few CMS platforms handle them awkwardly. Our rule: serve WEBP on your own website (convert with PNG to WEBP), but when sending an image to another person, JPG or PNG is the safer courtesy.

HEIC: great compression, constant friction

If you own an iPhone, your photos are probably HEIC. Apple adopted it in 2017 because it compresses roughly twice as well as JPG at similar quality — real money in iCloud storage across a billion devices — and it handles modern camera features like burst sequences and depth maps in one container.

The problem is that almost nothing outside the Apple ecosystem opens it happily. Windows needs a paid codec for full support, plenty of websites reject HEIC uploads outright, and Android support is patchy. So the format is excellent and the experience of sharing it is not. When a HEIC file needs to leave Apple's garden, convert it to JPG first; the recipient will simply see a photo instead of an error message.

SVG: not pixels at all

SVG is the odd one out because it is not a grid of pixels. It is a text file of drawing instructions — "circle here, this radius, this fill". That makes it a vector format: the browser redraws it at whatever size is needed, so one small SVG file is perfectly crisp on a phone, a 4K monitor and a printed poster alike. For logos, icons and simple illustrations, nothing else comes close.

What SVG cannot sensibly hold is a photograph — there is no reasonable set of drawing instructions for a sunset. And some places (Office documents, many upload forms, most chat apps) will not render SVG at all, in which case rasterizing it to PNG at the size you need is the practical way out.

A practical decision list

Mistakes we see constantly

Saving photos as PNG. Covered above, but it is the number one cause of "why is this file 20 MB". PNG is not "better quality" for a photo that already came out of a lossy camera pipeline; it is just bigger.

Re-saving JPGs over and over. Each save is a fresh round of lossy compression on top of the last one's damage. Edit a JPG once or twice and it survives fine; pass a meme through twelve apps and the mushy, blocky texture you end up with is generation loss. Keep an original and export from that, rather than editing exports of exports.

Upscaling raster images. Enlarging a 200-pixel logo to 2000 pixels does not create detail — the software just invents blur between the pixels you had. If you need something bigger than the original, go back to the source, or recreate it as an SVG. Scaling down is always fine; scaling up never really is.

For more short, situation-specific advice, our conversion tips page covers favicons, batch conversions, data formats and other everyday cases.