WebP vs AVIF vs JPEG, Measured on One Phone Photo

The short answer

On our test photo, AVIF made the smallest file at the same quality, about a quarter under JPEG. WebP landed in between, 7 to 13 percent under JPEG. JPEG made the largest file, and it opens in every major browser, old or new.

AVIF pays for that in time. Convert to AVIF needed about six seconds for the 12-megapixel photo at its default setting. Convert to WebP and Compress Image needed about 1.6 seconds each. For screenshots, lossless WebP kept every pixel and still cut the file by more than half. For transparency, pick WebP or AVIF: JPEG has none.

How we measured

On October 20, 2026, we took a public-domain photo from Wikimedia Commons. It shows a cliff and forest in Gifu, Japan, shot on a Pixel 9a. It is 4000 × 3000 pixels and 4.4 MB, a typical phone JPEG with sky, leaves and rock.

We dropped it into each tool on pulsarkit.com, in Chrome 155 on a 2023 MacBook with 12 processor cores. The clock ran from dropping the file, or from changing a setting, until the Download button turned on. The times below come from repeat runs. The first run of a visit is slower, because the encoder is still arriving.

Default settings do not give equal quality, so we also moved the quality sliders. To compare, we used SSIM, a standard score of how closely one image matches another, where 1 means identical. We computed it ourselves on the downloaded files, against the original photo.

Compress Image keeps the format, so our JPEG files are the phone JPEG re-encoded by MozJPEG. It uses quality 70 on Balanced and 85 on Max quality. A different JPEG encoder would give different sizes.

File sizes at matched quality

First the defaults, since that is what you get without touching a thing. Compress Image on Balanced turned the 4.4 MB photo into 1.5 MB. Convert to WebP at quality 80 gave 2.1 MB, and Convert to AVIF at quality 60 gave 1.7 MB. The JPEG wins this round only because it strays furthest from the original. Its SSIM is 0.957, against 0.977 for the WebP and 0.975 for the AVIF.

Matched to the JPEG's 0.957, WebP needed quality 60 and AVIF quality 50. The files came to 1.5 MB for JPEG, 1.4 MB for WebP and 1.2 MB for AVIF. That puts WebP 7 percent under the JPEG and AVIF 24 percent under it.

Matched at about 0.978, JPEG needed Max quality, WebP quality 81 and AVIF quality 62. That gave 2.5 MB, 2.1 MB and 1.8 MB: WebP 13 percent and AVIF 26 percent under JPEG. In both rounds, AVIF came out 15 to 18 percent smaller than WebP.

These are the numbers of one photo, and a picture with more texture or more empty sky will move them. Each tool shows the size before and after, read from the finished file, so checking your own photo takes seconds.

Encoding time in the browser

Compress Image and Convert to WebP finished the photo in 1.6 to 1.7 seconds. Convert to AVIF took 5.7 to 6.2 seconds at its default setting. The first AVIF run of a visit took 6.7 to 7.9 seconds. It also downloads the encoder, about 1.1 MB.

In the tab, the AVIF encoder runs on a single processor core. Above 6 megapixels, the tool also speeds itself up to keep the wait shorter. Our 12-megapixel photo was encoded at speed 7, though the Encoding speed menu still shows 6 as the default.

Setting Encoding speed to 10, the fastest option, brought the photo down to 1.4 to 1.6 seconds. The file came out 1 percent smaller, and SSIM moved by less than 0.001. On this photo, the fast setting cost nothing we could measure. On yours, compare the sizes the page shows.

Detail, screenshots and transparency

At 100 percent, we could not tell the three files apart at either quality level. At 300 percent and the lower level, each format leaves its own trace. The JPEG shows faint square blocks in the flat sky and along the treeline. WebP and AVIF keep the sky smooth, but they also wipe out the fine grain of the camera sensor. AVIF smooths the most: the dark gaps between leaves come out a little softer and flatter.

Small colored details suffer in another way. We zoomed in 8 times on a screenshot. In lossy WebP and in default AVIF, the thin red arrows of an icon turned darker and smeared. Both keep color at half resolution by default, and lossy WebP always does. Lossless WebP kept the arrows exact. AVIF with Colour detail set to 4:4:4 kept them clean, for 3 percent more bytes.

The screenshot itself, a 1280 × 800 PNG of a web page at 79 KB, favors lossless compression. Lossless WebP made it 33 KB with every pixel unchanged. Lossy WebP gave 31 KB and AVIF 27 KB, with the smearing above. Compress Image kept it a PNG and reached 25 KB by cutting it down to 122 colors.

For transparency we used a 600 × 200 PNG logo with soft edges, 9.4 KB. Convert to WebP made it 4.4 KB and Convert to AVIF 4.5 KB. Both kept its 206 levels of transparency, the same in every pixel. JPEG has no alpha channel, so transparent areas must become a solid color. When a transparent WebP becomes a JPEG, WebP to JPG puts white behind it. Another color can be set under Background for transparent areas.

Where each format will not open

JPEG is the safe one: MDN lists support in every version of Chrome, Edge, Firefox, Opera and Safari. On caniuse.com on October 20, 2026, WebP covered 97.18 percent of global browser use. AVIF covered 96.35 percent, partial support included.

The gaps sit in older software. Safari on a Mac shows AVIF fully from version 16.4, and only partly in 16.1 to 16.3. Edge shows it from version 121. Internet Explorer opens neither format, and Opera Mini opens WebP but not AVIF. WordPress accepts AVIF from version 6.5, and only when the server's image library supports it.

Google Search supports both on web pages: its image guidelines list WebP and AVIF next to JPEG and PNG. One difference shows on slow connections. AVIF cannot render progressively, so the browser shows nothing until the whole file has arrived. If a form or older program refuses WebP, WebP to JPG turns it into a JPEG in the tab.