I've just seen them for the first time. They work really well together. I hope they do more around image compression going forward. Considering the impact images have on page load times and therefore user experience, this type of content is very important. Surma and Jake are ideal evangelists. Google we need more of this.
This is incredibly informative! I knew quite a lot about image compression and graphical formats, but knowing the mechanics of what is behind it will really help me to plan ahead and get the most out of the images I edit and graphics I create. Cheers!
One caveat about the advice of "don't zoom in", if you're not on a retina display, it may still be worth zooming in 2-3 times as much as you want it displayed on the site to ensure it will still look crisp on high-dpi devices.
Yeah, things get tricky if you don't have a representative device to test on. But I'd still say that 'zoomed out to display size' is more representative than zoomed in. By zooming in you're making the artefacts physically larger than they'll actually appear to the user, which makes them easier to see
Fallen in love with squooshing, thanks to creators for sharing the intent, how it was conceived and how to get the best out of it. Its getting the 80% of the job done with 20% of the effort without becoming a coding or photoshop expert. Also had fun watching this quirky webinar.
Huge thanks from a zoomer! Understood, I'll stop zoom in my images, now... I'm campaigning for years for developers to reduce their images' size and this video actually has all you need to understand how compression works :D Image compression is important for UX because it reduces loading time, so users are happier, but what it also does is that it reduces the consumed bandwidth, which means less electricity used to carry the data, less servers' cpu load, so less servers' cooling, less disk space used, so less need to build hundreds of datacenters just to hold useless data, and so forth. Of course, this is not at the scale of a simple website that it will bring about major change. But who knows? What if every frontend developer were really picky on image file size? Would it reduce the average bandwidth consumed daily at world scale? Maybe not if users continue to consume this much videos - which leads me to think that all efforts I've put on image compression 'till now have been wasted by me just watching a 1080p 60fps video about image compression. I would love the same kind of video for video compression, by the way. Also that AVIF format is reeeaaaaally interesting.
My old boss would complain about the slightest artifact of compression, also whatever encoder photoshop uses is the best... Finally wanting transparency on photos for web... so all one can do is PNG24 save from photoshop with many, many colours. (Some of this was before AVIF and other formats that support transparency with lossy compression).
@@TsunaSawada26 hah you've just got good eyes then! It was tricky trying to come up with the right image for that bit - trying to find something that would work on most people but still be kinda obvious when it's pointed out. I tested it on about 10 people before I used it, and yeah, one of them also saw the noise first.
Jake Archibald nice nice i'm not trying to argue but the hilarious part is that I am wearing eyeglasses 🤣🤣🤣 maybe because when I looked at the picture, I focused on the weird part which is that noise 🤔🤔
Hi, when scaling down images, sharpness is lost. A large image with 25% quality setting is sharper than a small image with 75% quality setting. However, I'm afraid that my google pagespeed score and seo rating will down (as pagespeed prefers correctly sized images)
Great Talk, impressive results, thank you very much. But how do we get rid of JPEG/GIF etc. in the near future? All the frontend-folks are used to do it like they done it for decades, because it works for them. So why would they change? Doing all this image compression settings decisions by hand sucks bad if you have to decide it for every single image. I think the only way to change this "i've done it so for years"-behavior is to fully automate it - so it is actually an benefit to use it. This may be implemented server-side via CDN, CMS or HTTP server module or as Photoshop-plugin (WebPShop) etc. Are there any developments/experiments in this direction - so that once implemented there are less steps to think of? Just like: "upload that HighRes RAW Image, the server will serve whats best for the client". (It's like getting every http connection TLS secured. Unthinkable a few years ago, until the LetsEncrypt guys have automated it (and you Browser guys kinda forced it)).
Indeed, we do this in the EWWW Image Optimizer plugin for WordPress sites. Automation really makes WebP accessible for folks who don't have the time to fiddle around with every single image!
This might be the only reason I havent made the switch yet - it just doesnt feel like a file-ending of an image. Hilarious how resilient the mind can be to change sometimes! 😂
Near-lossless WebP coding can be a more powerful and generic option than going for 256 color palette. One needs to run it either a value of 40-80 (actually there are only three settings in that range, 40, 60 and 80). Values below 40 don't really work in a way that makes sense.
I definitely encourage folks to give it a try, but I'm not often happy with the results. Eg, try it with the Squoosh logo on squoosh.app, it creates 'halos' around some of the edges which are pretty noticeable. It might work for some images though.
I gave it a spin with Steve's 'team' image, and with slight loss at max the output is 24.5k. The visual looks fine, but still much larger than the palette reduced version. But yeah, it might work for other images.
@@jakearchibald In general the 'slight loss' is less aggressive at normal settings than palette loss. However, at maximal settings I suspect that "slight less" is overly aggressive. Do you know how the slight loss maps to cwebp's -near_lossless values?
@@jakearchibald then run at slight loss = 40 or slight loss = 60, but yes palette reduction will be more compression but also less predictable in quality.
We have an eCommerce site with hundreds of product photos. Obviously, we can't hand week each photo. Currently, we upload them as png files. We have a class there that generates jpg files in various sizes for the site to use (and caches them). What is a better solution for this situation?
You can use an opensource lib for webp or use a caching service like Cloudflare etc.. Also, i would convert them to JPEG first then send them off to the Server for thumbnails etc.
Why Google employees don't use Chromebook, everyone use MacBook, even in the summits. And they say others to use Chromebook. Dagger in the heart of marketing
I use a Chromebook for general browsing, and to watch things like Netflix on the go. However, I use a mac as my development machine, as you can see in this video. I don't think that 'stabs' any particular marketing strategy.
Great talk as always. The Zoomer joke kinda threw me off cuz it was so unexpected. Unrelated question tho, are you planning on updating PROXX in the future? Cuz there are some improvements that can be made to make the game more enjoyable (shameless self-plug: I submitted a PR with a couple of changes that I thought would be useful lmao, so check that out if you want to)
How do we do the compression in real-world applications? Is there an open-source algorithm I add to my script where I just enter my image in the argument of the function and a webp image is returned? Or do I have to use some library?
@@jakearchibald Awesome. I'm sold on webp watching all these videos and after testing it myself. Was wondering if there was a simple way to include it client-side in my svelte project.
Is this open-source? I'd love to add something like this to my market company site. It's a bit of a hassle and some people always seem to forget the compression step. I understand the compression aspect no problem, but the application of this on the front end is crazy. I'd almost have to guess that it is hitting a backend api on the preview side in order for it to compress and show you the data. Otherwise, Id really love to know how the data was calculated. I've struggled finding a way to do that on the front end. Is mozJPEG a api compression service? Is that how this works? I have so many questions. Somebody help lol
@@jakearchibald oh wow thanks man! This is really cool I'm excited to see how yall did this, so I'll have to check it out. At first id have guessed maybe it was serverless, but I fell into a Web Assembly rabbit hole after watching this yesterday, and this seems like the perfect application for it. I'm hype to find out. Thanks again for this man. Great work from you and your team
Real users are going to zoom in like that. This is the digital world where people rely on eCommerce for personal and professional procurement. Webp looks absolutely horrible and destroys brand perception for eCommerce. One would think that Google would come up with something better/other than this and not degrade ones web page score because they don't use your "make-my-picture-95+%-worse-looking" applications. Appreciated effort. Unappreciated name calling and total misuse of brand identity. Fail, Google.
Some users will zoom in, but does that mean you should serve all images at 4000x4000 just in case, making things slow for everyone? I don't think so. We have the tools to detect when users pinch-zoom (the visual viewport API), and that seems like a better cue to load in higher quality, higher resolution imagery. I don't agree with your statements around WebP, but as I've pointed out in another comment, we have AVIF now, which performs even better.
The fact that the squoosh stuff is possible is nice, but its not practical. What we need is an npm package and/or webpack plugin that automates this process. Manually doing each image is not practical.
Right, but you can use Squoosh to find the optimal settings for the type of image you're expecting. Having said that, a lot of images, especially those related to site design, don't often change, so manual tweaking there is fine.
"WebP supports both lossy and lossless compression; in fact, a single animation can combine lossy and lossless frames. " From the section *"How do I use libwebp with C#?"* scroll down below the code for the headline "Why should I use animated WebP? Advantages of animated WebP compared to animated GIF?" developers.google.com/speed/webp/faq
Guys - can Squoosh have some form of help or explanation. I don't need the science, I've Google for that. Just a few lines explaining sliders and tick-boxes.
Using the same WebP container for both lossy and lossless content is in my opinion a terrible design decision. It's obfuscating information that's very important to a lot of people for no good reason, making it very cumbersome to find what you're looking for when you're interested in just one of the two. That slight loss option makes it even worse. Imagine looking for lossless images and finding a WebP and figuring out that it's with the lossless codec, but then it's actually not lossless at all and you have no way of knowing until you start working with it and zoom in or whatever (which may be very important for the use case).
Doesn't PNG have the same problem? If you find a PNG, you don't know if the palette was reduced before it was encoded. I also kinda wish that webp and webp-lossless were kept seperate, but then there's a long history of this with audio and video codecs. "wav", "mp4", "mkv", "avi" - all of these container formats can contain various different codecs.
Yes, of course you can never prevent lossy preprocessing, so the responsibility to keep things clear and separated falls on the encoders. If you select that you want lossless as a base, then there really shouldn't be any secondary configuration that makes your lossless compression... well, not lossless anymore. The only time I could see it actually being relevant is if you for some reason only can decode the lossless codec and still want lossy compression. That seems far-fetched though.