Google Is Already Experimenting with WebP2 as Successor to WebP Image Format | Hacker News Reader

Google Is Already Experimenting with WebP2 as Successor to WebP Image Format

phoronix.com

45 pointsvarbhat59 comments

Maybe we can slow down development a little bit and actually design a format that will last for another 30 years instead of requiring browsers to support yet another hot new format every five years.

why bother when you can beta test on the most popular piece of software in continuous use around the world?

I think that ship has sailed a long time ago... software needs regular updates for security patching, so adding new formats is not the big deal it once was...

If nothing else, just provide a slow-but-correct decoder that translates the format to a long term stable format (e.g. PNG).

An in-app decoder would translate to pixel data. Why would you re-encode to a different format like PNG that's optimized for storage when rendering to a display?

Sorry I meant recode to pass into tools/libraries that no longer support the older format.

That doesn't work so well with the whole ecosystem around it that actually creates those files. Think of graphics programs that still don't support WebP or cameras that cannot simply be updated... oh wait, smartphones already killed those.

Why does it matter if a camera supports WebP or not?

Because nobody actually needs to use this format if it doesn't benefit them. If the headline was "Google developing new proprietary image format to reduce web traffic to Google services by by 2% in Chrome" it would be a non-issue.

It seems odd that Google is being punished for releasing their work to anyone who wants it.

If it was framed that way, they’d be wide open to accusations of anticompetitive behaviour—using their position of operating their various web services and making a browser to push functionality that benefits just them into their browser, which no other company would be able to do.

That paints kind of a weird perspective on open source and open formats. Companies develop and release lots of OSS specifically because it benefits them to write and maintain it. Twitter made twemproxy for their own scaling issues but then released it to the world. I don't see a difference with Google here.

"We made this thing because it solves a problem we have, and we're giving it away so it might solve the same problem for you" doesn't seem nefarious.

This is one of the many problems that come from Google having so much market power. This is really the Google development practices being forced onto everyone; iterate fast, create new stuff and throw old stuff away when you feel like it.

I do hope other browser makers like Mozilla Firefox and Apple Safari will hold back on this, like they should have done with Webp v1.

Safari only just got webp support super recently. Seems like they held back on it long enough...

Also note that it’s only Safari 14 on the new OS releases; Safari 14 on Mojave can’t view them.

WebP was introduced in September 2010, Mozilla started supporting it in Firefox in 2019, and Apple is just adding it to Safari now. You think they should have waited longer than 10 years?

(Disclosure: I work for Google, speaking only for myself)

They should have waited indefinitely. WebP has been a big source of confusion for non-technical people I know, trying to download images from the internet.

If you want to stop progress for the least technically inclined to accept you will stop all progress for ever.

> If you want to stop progress for the least technically inclined to accept you will stop all progress for ever.

This is a strawman, at no point am I or would I ever claim this is a good idea.

At what point is progress too much progress? Why can't we also balance these things with backwards compatibility, interoperability, accessibility, and other such things?

> least technically inclined

How many people actually know what WebP is, as opposed to thinking its some sort of bug with google or their browser? The least technically minded are more than a majority of users, more like the vast majority. Is their usage not valued?

Once WebP starts to get used more and more and people will stumble on issues with *.webp files they have no idea what to do with, their OSes and tools that they use will start to handle WebP files and the users will eventually learn what it is the same way they learned what JPEG is in the past. I don't know where the line of "too much progress" is, but you need to start that feedback loop somewhere.

Yes, I also think they should have waited indefinitely. I don't trust Google to care about backwards compatibility and keep on supporting Webp v1 forever, I would expect them to drop it before even dropping jpeg or png. Especially in this case, where it mostly serves Google and other big web corporations, and not regular users.

Mozilla always had the stance that they wouldn't just add support for technology that might be just a passer-by technology. I think that was the right choice and I think it is a shame they added support for webp recently.

I can even imagine this to be market practice, where they add new stuff, where Apple and Mozilla and newcomers have to chase Google's fad of the day, while Google has enormous resources for Chrome. You can call me paranoid, I probably am :)

Anyway, if Google Chrome ditches Webp v1 in the next 10 years, I will not be that guy to say "Told you so!", but I will think it for sure.

Nah, why would Google do that? Oh wait... that’s exactly what they’re doing with HTTP/2 Push.

Server push is an optional extra, and the browser can drop support without hurting compatibility at all.

The browser cannot drop support for an image format without major compatibility issues.

If you would like to bet on whether webp is still supported in Chrome in a decade I'd be happy to ;)

(Disclosure: I work for Google, and have no internal information on this)

Not going to happen as long as people keep following their lead. It's in Google's interest to do things like maximize compression (especially for video and images) since it reduces their operational costs in the form of transmitting fewer bytes. Sure it benefits everyone (a bit), but while others might prioritize a bit more stability in standards, Google (and other major hosting services) value the reduced opex now. For most of the rest of us, we're well into diminishing returns on chasing the latest codec.

At Google, you can't get promoted by steering the ship.

Do you think that image (or video since a lot of tech overlaps) formats move so slowly that over 30 years there is no innovation worth considering?

I'm pretty sure feature churn and foisting technical debt onto your competitors is a feature, not a bug for Google.

JPEG XL is the next format that I am sure will take over:

https://cloudinary.com/blog/how_jpeg_xl_compares_to_other_im...

It's virtually lossless and progressive.

After 400 re-saves it's visually unchanged: https://twitter.com/jonsneyers/status/1327302393776758784

JPEG XL looks impressive, but it's still under heavy development. It might eventually take over, but AVIF is already starting to have browser support: https://caniuse.com/avif

and then you just have to wait 20 years for all the submarine patents to expire. So sometime around 2040-2045 we will get a better format.

Any format can be subject to submarine patents, even the ones that require licenses.

Not when they are already in use and had all patents expired. Example, MPEG 2, MP3 and AAC-LC

It is not very good at the low bitrate ( compare to State of the Art ), Although the author did said they spend very little time in optimising.

And Still image with VVC at low bitrate is even more impressive.

We need it to be both smaller and better than the best of JPEG. Which is still getting many improvement.

[one of WebP2 author here]

This 'new' WebP format is not radically different, but just pushes the same use-cases further (The web. Not archival. Not camera capture.), sticking to simplicity and usefulness.

10 years is a long time since WebP-v1, and techno has evolved since. We can go much lower in bitrate for still-good-enough quality, and that's what we're interested in exploring.

AVIF was designed starting from a video codec, similar to WebP-v1 in the early stages, and this has some drawbacks (that later we needed to correct in WebP, that was confusing). WebP2 is aiming at images-on-the-web right away, and not as an afterthought.

More details about the last 10 years of WebP and some WebP2 features: http://shorturl.at/BEMV8

The responses in this thread are disappointing. The idea of chastising a person or organization for iterating or creating a new format is something that I can’t agree with

Maybe this will be a future popular format. Maybe it'll die on the vine. Who knows, but experimentation and alternate approaches, particularly when the code is being shared is always a good thing, regardless of how you feel about Google or Chrome

I thought the issue leading to this was AVIF decoding complexity but the goal of this project is apparently roughly parity with AVIF decoding performance. Is encoding performance of still images such a big deal?

I thought these problems were largely solved if they just gave implementations time to mature. Unless we’re talking wildly different concepts here.

I guess I will need yet another new add-on in Firefox to download images as PNG or JPG (as I currently need to do for WebP).

What extension do you use for this? I had an issue the other day with this, which was a bit annoying - FF wanted to save an image as webp rather than a regular jpg, for some reason, and in the end i was forced to use WGET instead.

Sure, I use this one https://addons.mozilla.org/en-US/firefox/addon/save-webp-as-...

And only 10bit HDR support? Weak. Though I guess PQ has been patented by Dolby?

https://www.avsforum.com/threads/smpte-webinar-dolby-vision-...

Can we please calm down until at least one app on my machine besides the browser supports this?

But these formats are basically for browsers. I don't think Google cares one iota if they gain traction anywhere else because having them used on Google Services and Chrome is enough to reap the benefits of x% less traffic.

Is it just me, or has Google become a corporate-AI version of a paperclip optimizer, but for money?

Perhaps it's not just Google, but this kind of "F[orget] the user" behavior is pretty concerning.

I don't get your last point. This is one of those situations where it's a win for Google, users, and any site that also adopts the format.

Google and sites that use it use fewer resources delivering traffic and sites get faster and smaller for users. Sure you could rant about all kinds of different web bloat but a completely transparent improvement is always a good thing.

I don't agree that it's a win for users, because the format can only be used in the browser. There may be work-arounds, but forcing users to work around a largely unsupported image format is hostile to users.

I mean it’s not only supported in browsers just the motivation for its existence is browsers. It will likely get the same level of support as WebP which is “not ubiquitous but well supported.”

If you are someone who looks at image formats and says “wow this is really wasteful, we could be doing so much better” then your constraints make it a no win. Nobody adopts formats that aren’t supported, and nobody supports formats that aren’t used. And users are just stuck with waste forever. Somebody has to break the ice and Google using the format in a non-disruptive setting of Google talking to Google is pretty conservative.

If it was Adobe’s new format they would bootstrap it by making it the default in Photoshop or Apple would make it the default format for the camera or something.

"The aim is to be around 30% better than the original WebP design but is looking about 20% worse than AVIF"

Why isn't WebP2 what WebP was? WebP was basically just a portion of the VP8 codec.

So why isn't WebP2 the same thing but for AV1? Why is WebP2 not just AVIF?

Browsers aren't required to do anything. Do you remember getting PNG transparency to work in IE6?

Progressively enhance the images you decide have enough business value to do so. Where I work we never replaced GIFs with APNGs, but we _did_ replace them with MP4s.

It's hard to get excited for the web. I guess Apple will support it sometime never, as they don't support webm in safari.

Apple supports WebP and VP9 as of iOS 14. But not WebM.

I've come across WebP images a few times in the past whenever I've needed to download images for marketing material, etc.

It's usually impossible to: import those images anywhere, modify them, or basically do anything with them (even open them outside Chrome).

As a consequence I've had to install another browser to ensure the images were not WebP.

Was that the intended purpose?

appending ?jpg to the url will sometimes get you the original jpeg file.

Failing this, browser extensions such as 'save image as you want' will do the trick.

https://chrome.google.com/webstore/detail/save-image-as-you-...

The purpose is to reduce traffic between Google services and Chrome. That use-case alone is enough for Google to justify the investment.

Thx, I'll stick to JPEG.

I hope you work in e-commerce, my manager is going to be thrilled to hear you're sacrificing money for dogma.

Maybe the Android team can bother to actually provide first class support instead of a github dump that we are supposed to integrate ourselves then.

History of replacing JPEG with something better

- PNG Mostly to replace CompueServeGIF (Not to be mistaken for Yiff files). Neat but no real gain over JPEG in quality on photos so nonone really used it for anythign but animations, but everyone were already used to GIF in their workflow so.. meh...

- JPEG2000 1997-2000 embroiled in patent nonsense so noone really uses it in the wild.

- WebP: used by google to look cool but since most people used a webbrowser that does not support it at launch servers were forced to have JPEG's as fallback. Great savings much wow.

- HEIF: Absolute cluster fudge of variations then Apple did a big Dick move and said "Orchard says HEIC, or Bust!" without actually telling anyone and broke several people's workflows that depended on JPEGS form iPhones, especially since FCPX did not support HEIC on launch and it took another 6 months for that update to come. Yay.

- WebP2; Google want to swing big dick dev powers to gain a marginal possible performance savings on their servers and say's "WebP? That is sooo 2015! you gotta be on WebP2 to hang with us"

Is it really that surprising that most people who depend on workflows to actually work to make their money are still using JPEG, GIF and TIFF files?

hehe, have you heard of AVIF? :D

PNG was never intended to be a replacement for JPEG. JPEG is for continuous tone images. PNG (and GIF) are optimized for line art.

PNG replaced GIF because the latter was patent-encumbered at the time, only supported 8-bit color, and in later revisions only support 1-bit alpha transparency. PNG supports 24-bit color plus an 8-bit alpha transparency channel.

WebP, first released in 2010, allows for the replacement of GIF/PNG with lossless encodings and lossy encodings for replacement of JPEG. And unlike JPEG, WebP supports alpha transparency as well.

10 years sounds like more than enough time to optimize previous work and publish it. It's not like existing WebP or even new WebP images will suddenly stop working.

PNG is now actually the most used format on the web: https://w3techs.com/technologies/overview/image_format significantly more popular than GIF.

BTW, WebP is used by 0.4% of all websites.

> broke several people's workflows that depended on JPEGS form iPhones,

Reply on news.ycombinator.com