Soon™? Google works on an optimized JPEG web image format named PIK

[LigH1L]

https://github.com/google/pik

The goal is to compress images more than 50% better than JPEG in visually lossless quality (freeze planned at 65% better compression efficiency), while decreasing decoding speed only less than 50%.

It uses a combination of two other Google projects with funny Swiss pastry related names: Butteraugli (HVS-aware image differences) and Guetzli (JPEG encoder with denser packing).

Current implementations don't require AVX2 and FMA anymore, but still expect SSE4; this CPU feature threshold may drop further during the development.

So, no hurry yet but keep it in mind, some day it will probably be supported in Chrome.

[Lootyhoof]

So...what was wrong with WebP?

[adesh]

This is how Google rolls! They create one thing, and then few years later something else for the same purpose. Remember Anguar1 and Angular2.
This is software pollution. They do not do it right the first time.

[Tomaso]

I wonder what types of nasty new tracking methods this will allow for them to use.

[Moonchild]

Abusing their market dominance again, I see.
It's the old Embrace-Extend-Extinguish strategy all over again.

And yes, what's wrong with WebP? It's actually finally seeing larger adoption and now they want to introduce something new yet again?

Off-topic:
PS: "pik" is a rather common word for "penis" in lower Dutch.

[Moonchild]

As an update:
I looked at the Guetzli compressor and it will spit out widely-accepted standard-format JPG images that are smaller/denser and would use the same phychovisual model as "pik" for compression.
Also, it doesn't seem "pik" would support transparency like WebP, meaning it doesn't overcome the problem with jpeg as a web image format compared to png and webp and would simply aim to replace jpeg, with all its limitations. I don't see that happening.

EDIT: Poked the devs with some question: https://github.com/google/pik/issues/33

[Tomaso]

Don't know much about programming, but I don't like the hits that comes up when skimming through their code, using search terms like "apicid", "cpuid" and "unique".
Perhaps you could enlighten us, Moonchild?

[Moonchild]

I have no intention of wasting my time on a sec/privacy audit of their code.