Converting WebP to JPG: quality and file size

What actually happens during conversion

A WebP image consists of compressed channel values. Converting it decodes those values, returning the data to a grid of pixels, and then encodes that grid as JPEG. The pixel dimensions stay the same: 1600 × 900 pixels remain 1600 × 900 pixels. Only the way those pixels are stored changes.

The converter does all of this in the browser. The file is read, decoded and written again without leaving your device. No upload request is made for the conversion.

The quality level is a compression level

JPEG does not store images pixel for pixel. The encoder splits the image into blocks, converts the colour values into one brightness and two colour channels, and quantises the resulting coefficients. The quality level controls how coarse that quantisation is.

A value of 90 % therefore does not mean that 90 % of the image information is preserved. It names a step in the encoder's quantisation table. Two images at the same level can end up with very different file sizes, because their content carries different amounts of detail.

Where the file size comes from

Plenty of high-frequency detail – foliage, fabric, noise, fine type – produces many coefficients and therefore more bytes. Soft gradients and large calm areas compress much better. That is why a size quoted for a quality level is always an observation on that particular subject, never a general rule.

When converting WebP to JPG, the output can end up larger than the source. WebP uses a different prediction and reaches more compression per visible quality on some subjects. In that case the converter shows the actual difference instead of claiming a saving.

Which level for which purpose

• 90 %: a good default for images that will be edited further, shared or uploaded.
• 80–85 %: sensible for large image galleries on the web, when file size matters more than the last bit of detail.
• 95 %: for prepress and archiving, when the file will be passed on anyway.
• 100 %: not a lossless mode, but the finest quantisation of the lossy encoder – the file becomes considerably larger without showing visibly more.

Avoid repeated re-encoding

Every JPEG encoding loses information, and those losses add up. An image converted back and forth five times looks worse than a single pass, even at a high quality level. If you are unsure, keep the original file and produce the JPG version once, at the size you actually need.

Limits of processing in the browser

Up to 30 files are processed per run, each at most 40 MB and 32 megapixels. These limits are practical: decoding and encoding large images uses memory that a browser tab only has in limited supply. Larger files are rejected with a message instead of blocking the tab.