WebP: how compression changes the size of images

Useful to know

WebP is an image format developed for the purpose of effectively presenting graphics on the web. It supports both lossy and lossless compression, and also transparency and animated images. This brings together several options in one format family. However, the .webp extension itself does not indicate what settings the file was created with. Two images of the same size and the same format can differ significantly in volume and visible detail. [1]

In lossy compression, some information is changed to reduce size. The Google documentation describes the use of predicting values in blocks and encoding differences. The idea is that the adjacent parts of the image often have similar characteristics and it is not necessary for each to be described independently. Stronger compression can save more bytes, but its cost should be sought in the visible result, especially in text, contours, and fine structures. [2]

Lossless mode pursues another goal: restoring accurate pixel content instead of allowing a visually acceptable change. Conversions and prediction are also used here, but the information must be able to be recovered. This does not mean that any such file will be smaller than any alternative. The result depends on the structure of the picture. A monochromatic icon, a screenshot and a photograph of a forest set different tasks over the algorithm because their repeatability and detail are different. [2]

Transparency is a separate useful feature. It allows the image to be displayed on a different background, without the object necessarily having an opaque rectangle. WebP can combine transparency and compression with color part losses. When converting, you still need to check the behavior along the boundaries of the object. An image that looks good on a white background may show unwanted edges on a dark one. This is a question of the specific outcome, not just the support of an alpha channel. [1]

Google publishes comparisons with other formats, but such measurements have conditions and do not constitute a guarantee for any photo. For own site more useful is a small representative sample of the actual images. The size of the files, the appearance at the real width of the display and the behavior in the browsers used are compared. It must also be taken into account whether the tool retains the required metadata. Changing the format does not replace checking the entire process for creating and publishing images. [3]

Let's take an article with a big picture and a little schematic with captions. A general setting may be acceptable for the picture and blur the letters of the scheme. It is better for the setting to follow the type of content. The appropriate file is the one that performs the visual task with a reasonable volume, not necessarily the smallest received variant. WebP provides useful tools for this choice, but the final judgment remains a comparison between the particular image, its use, and the quality the reader actually sees.