Firefox 92 Beta Takes Flight With AVIF Image Support

[ssokolow]

Google is only responsible for the (VP9-derivated) compression modes of webp, the container format itslef is standard RIFF as used since Windows 3.1(for WAV, AVI, RMI, etc.) (basically like Amiga's IFF, but little endian)

That's part of why it's such a terrible format. It's a container that can hold lossless or lossy content, which automatically makes the lossless variant inferior to PNG since, with an animated GIF, you could wind up shocked or grossed out but, with WebP, you have to make sure it's not a lossy transcode that you're archiving.

It's everything that PNG explicitly rejected in pre-emptively making any attempt at APNG a spec violation, but worse. (The PNG spec says that, to be spec-compliant, a file with a PNG header must contain only image data derived from a single source pixmap, defined as a rectangular array of rectangular pixels, and transformed by a list of transformations that does not include splitting frames out of a spritesheet-like form. That's why Mozilla still has to maintain their own fork of libpng.)

Can you even detect that programmatically without a full-blown WebP parser? I know APNG can be detected using fairly simple code suitable for something like "send non-spec-compliant user avatars for conversion".

[Quackdoc]

I've recompressed 24MP JPEG photos made by my camera to around 40% of their size with zero perceivable quality loss and I'm not sure how Webp can be considered useless after that.

Did I say useless? No. I said mediocre. which it is. the average file size reduction from jpeg to webp I believe was around 30% where as avif is 50%. webp is also fairly okay, but not great for animations too.

[quikee]

I've recompressed 24MP JPEG photos made by my camera to around 40% of their size with zero perceivable quality loss and I'm not sure how Webp can be considered useless after that.

I'm sure that re-compressing those JPEG images again with a JPEG encoder will make them smaller, because the images weren't pushed to the perceivable limit. In tests with a better JPEG encoder like MozJPEG the difference is around 10% and even MozJPEG winning in some situations (VP8 destroying detail). Sure WebP doesn't fall apart as quickly as JPEG when you lower the target filesize, but it has some major flaws, like supporting only 4:2:0 subsampling (even JPEG can do 4:4:4) and magenta cast in pictures even, when you allow the encoder to use way than enough bits (probably a flaw in encoder, but there is only one WebP encoder so there is no other choice).

edit: If they would switch from VP8 to VP9 for WebP when they had the chance, it would be a much better image codec today. VP9 fixed a lot of issues that were present in VP8 and it would allow for 4:4:4 subsampling too.

[Daktyl198]

Did I say useless? No. I said mediocre. which it is. the average file size reduction from jpeg to webp I believe was around 30% where as avif is 50%. webp is also fairly okay, but not great for animations too.

Worse. The Mozjpeg encoder can create standard jpeg-compliant encodes that are ~8-12% on average bigger than a webp encode (confirmed with my own testing) with similar visual quality. As it creates standard jpeg files, any and all jpeg decoders can use the files, including libraries meant to modify them/etc. So in reality webp is only around 10% better than jpg in terms of size reduction, while avif is more like 20%.

The real winner in this race is JPEG-XL, though. It's not based on AV1, but is still aimed at being a free file format and absolutely destroys the competition.

[mavit]

Right-click menu doesn't list Gimp as an open-with target despite GIMP supporting it, you have to launch GIMP manually and then open the image from the File->Open menu.

That would be a bug of GIMP's package (or its installer) for failing to register the type into your desktop environment.

I had a look into this last year, and it turned out that it was caused by a disagreement about whether the content type was image/webp or image/x-webp. Should be fixed in the latest GIMP, even though it wasn't really GIMP's fault. This does highlight that getting a new image format widely adopted will involve lots of little niggles, though.

[cl333r]

I had a look into this last year, and it turned out that it was caused by a disagreement about whether the content type was image/webp or image/x-webp. Should be fixed in the latest GIMP, even though it wasn't really GIMP's fault. This does highlight that getting a new image format widely adopted will involve lots of little niggles, though.

Do you happen to know why would they need to stick in the "x-" in "image/webp"?

[quikee]

Do you happen to know why would they need to stick in the "x-" in "image/webp"?

The MIME types need to be registered and approved by IANA, and otherwise they should use the unregistered names i you don't plan to register the type, which use the "x-" prefix. However, before a type is registered it is a bit of a grey area, if it is OK to use the registered name or it should need to use the unregistered one and switch to the registered one once it is approved.