Compress WebP

Compress WebP: what our nine test files saved

We ran nine WebP files through this tool on 11 November 2026; the table above lists every result. Six we made ourselves from our own test images: two photos (one saved at quality 90, one at quality 60), a screenshot, a product cutout, a logo and a three-frame animation. Three came from real websites: a photo from Google's WebP Gallery, an image with transparency from Google's Lossless and Alpha Gallery and an image from the PixVibe blog.

The tool always saves the colour data of a WebP in lossy mode at the quality you set (transparency is kept lossless); there is no lossless WebP setting. The screenshot and the Google Lossless and Alpha Gallery image, both lossless, shrank 88% and 87%; the logo, also lossless, shrank 36%. The photo saved at quality 90 shrank 59%. Three other lossy WebPs shrank less: 35% for the blog image, 13% for the product cutout and 8% for the Google WebP Gallery photo; the photo saved at quality 60 did not shrink at all and was handed back unchanged, labelled "Already small enough, kept as is" (in a batch, the card shows "Kept as is").

In Compare with original, the quality-90 photo and its 64.3 KB result looked the same at the size the viewer opens with; zoomed to 338%, fine skin texture and loose hair looked slightly smoother. The sneaker cutout at 225% showed no difference we could see in the mesh, laces or grey flecks. The screenshot, a lossless file saved again as lossy, at 338%: the text stayed sharp and readable, and the thin light-grey outlines around small buttons looked slightly softer. The logo at 225% showed faint blotches along the edges of its dark lettering that are not in the original.

How we measured: Chrome on a Mac, By quality with Quality at 75, and Output format set to Original format for the WebP rows and to JPG for the JPG rows. Every size is the file's byte count divided by 1000 (1 KB = 1000 bytes), rounded to the last digit in the table. The tool's own display rounds up, so its batch summary for the WebP run read 2.30 MB to 1.06 MB while the bytes add up to 2.29 MB and 1.06 MB. Those totals include the animated file, which lost two of its three frames; without it the bytes add up to 2.18 MB and 1.03 MB.

Transparency and animation in a compressed WebP

WebP can hold more than a plain picture. The IETF's RFC 9649, WebP Image Format (2024) describes WebP as a format "that supports lossless and lossy compression as well as alpha (transparency) and animation". This tool keeps the first of those extras and drops the second.

Transparency: the three test files with transparent areas came back with a separate alpha chunk, and when we compared them pixel by pixel, all 1,572,864 pixels of the sneaker cutout, all 992,640 of the logo and all 120,400 of the Google Lossless and Alpha Gallery image had exactly the same opacity as before, including its 1,009,008 partly transparent pixels (opacity 1–254 out of 255; in this cutout even the shoe itself is not fully opaque). We repeated the check on the cutout compressed to a 600 KB target size (result: 598.9 KB), and again every pixel's opacity matched.

The cutout also shows why a WebP with transparency can be hard to shrink. Its transparency layer is stored in a chunk of its own, which took 472.7 KB of the 774.4 KB file. This tool saves that layer losslessly, so it came back at 474.2 KB; only the colour data was saved again at quality 75, falling from 301.7 KB to 199.6 KB. In our file, most of the bytes were transparency that the tool does not compress.

Animation: our three-frame animated WebP came back 79% smaller because only the first frame was kept; the result is a still WebP. The tool shows no warning when this happens. If the animation matters, keep the original file and use a tool built for animated images.

WebP or JPG: should you change the format?

Changing Output format to JPG did not reliably help. For the lossless screenshot, JPG at quality 75 gave 198.2 KB against 118.0 KB as WebP; for the logo, JPG gave 30.6 KB, which is bigger than the 24.0 KB original; and the photo saved at quality 60 grew from 55.5 KB to 94.6 KB as JPG. The rule that a file never comes back bigger only covers WebP output, so the tool hands you the bigger JPG in those cases.

In seven of our nine files, the WebP output was smaller than the JPG output. The two exceptions both had transparency, and JPG only won by dropping it: the product cutout was 152.9 KB as JPG against 673.9 KB as WebP, and the Google Lossless and Alpha Gallery image 10.5 KB against 10.7 KB. In both JPGs the transparent areas turned white.

Google's WebP page says: "WebP lossless images are 26% smaller in size compared to PNGs." That compares lossless WebP with PNG, not with JPG. Our one screenshot happened to land at 26% too: 1.29 MB as PNG and 952.2 KB as lossless WebP. Switch formats only when the place you upload to does not accept WebP; for that, use WebP to JPG or, to keep transparency, WebP to PNG.