Every CMS, Joomla, WordPress, Drupal, or anything else that lets users upload images through an admin panel, eventually has to answer the same question: which modern image format should it use to generate thumbnails and web sized images. AVIF and WebP are the two candidates on the table, and AVIF often gets pitched as the future because it can compress harder than WebP on paper. When it comes to a CMS that generates images on its own server, that promise doesn't hold up well in practice. Based on the evidence below, WebP looks like the more practical choice for CMS generated images today, while AVIF is not yet worth building into a default image pipeline.
AVIF isn't a format designed from scratch for still images. It's a still frame container built on top of AV1, a video compression codec developed for streaming video. That heritage matters, because video codecs are built and tuned around a specific workflow: encode once, offline, on dedicated hardware or heavy server infrastructure, and decode billions of times afterward on cheap client devices. That trade off makes sense for a streaming platform, where content is encoded once and served forever.
A CMS works differently. It needs to encode images on an ongoing basis, on the server, at the time a user uploads a photo, or whenever a theme changes a thumbnail size, or when a gallery gets rebuilt. Applying a video codec's still image mode to that kind of workflow means paying video-encoding style costs on every single upload, on whatever shared or mid tier hosting the site happens to run on. That mismatch is worth keeping in mind when comparing the two formats for CMS use.
When generating images on the server using the GD library at 100% quality, the following was measured across formats:
| Format | File size | Server generation time |
|---|---|---|
| PNG | 7.04 MB | 0.10 s |
| JPG | 2.10 MB | 0.03 s |
| WebP | 981 KB | 0.33 s |
| AVIF (default speed) | 3.72 MB | 1.43 s |
| AVIF (fast) | 3.87 MB | 0.39 s |
| AVIF (slow / high compression) | 2.83 MB | 46.70 s |
Two things stand out here. First, at default encoder settings, AVIF didn't even beat WebP on file size. The compression advantage AVIF is known for only shows up once the encoder effort is turned way up, since that's how a video codec finds extra savings: by spending much more compute searching for redundancy. Second, turning that effort up moved generation time from a fraction of a second to 46.7 seconds for a single image. For a gallery with hundreds of photos, or a CMS that regenerates thumbnails whenever a template or size setting changes, that kind of cost adds up quickly and can turn into hours of server time, or a request that simply times out.
This lines up with what broader image format benchmarking reports in 2026: AVIF encoding is commonly measured at somewhere between 5 and 50 times slower than WebP, and can add several seconds, sometimes considerably more, per image during upload processing. It's part of why performance focused CDNs tend to run AVIF encoding as a background job rather than doing it inside a live request. A CDN can be built around that kind of infrastructure. A typical CMS install on shared hosting usually isn't.
A generic AVIF vs WebP comparison often assumes a build pipeline: images are converted once, pushed to a CDN, and never touched again. A CMS doesn't usually work that way, and several practical points follow from that:
None of this is meant as a broad dismissal of AVIF as a technology. It's simply what tends to happen when a codec built for offline video encoding is applied to a live, per upload image pipeline running on ordinary web hosting.
WebP was designed from the start as a web image format rather than adapted from a video codec, and that shows up in how it behaves in the tests above and in general use: encoding stays fast and fairly predictable regardless of how the compression settings are pushed, support in GD, ImageMagick, and CMS plugin ecosystems is broad and mature, transparency is supported well enough to replace PNG, animation is supported well enough to replace GIF, and browser support sits around 97% or more of global traffic, largely stable since Safari added support in 2020. There's no equivalent to AVIF's silent mismatch failure mode, no need to check whether a given host's ImageMagick build is new enough, and no realistic scenario where uploading a single photo holds up a request for tens of seconds.
Whether it's Joomla, WordPress, or Drupal, the evidence points in the same direction:
AVIF's compression numbers look strong in isolation, but they come from a codec built for a different workflow: encode once offline, decode indefinitely afterward. A CMS operates closer to the opposite pattern, encoding on an ongoing basis, live, on hosting environments that vary a lot in what they support. That difference is a large part of why AVIF runs into slow encoding, inconsistent server library support, and quiet failures in exactly the setting a CMS operates in, while WebP, built with the web in mind from the outset, tends to just work. Based on the current evidence, WebP is the more practical choice for CMS generated images.