PICTURE PERFECT A D D Y O S M A N I MODERN IMAGE OPTIMIZATION

Video by KoolShooters from Pexels PERFECT PICTURES ON THE WEB

No content

Optimizing images. How hard could it be?

Video by fauxels from Pexels IMAGES ARE KEY

Image optimization is an evolving topic.

Major changes in the image landscape • WebP support in all modern browsers • AVIF image decode support landing in Chrome, Firefox and Safari • JPEGXL decoding behind a fl ag in Chrome • Lazy-loading images in Chrome, FF, Safari. A default in WordPress • Compute img/video aspect ratio from width/height attributes • FF support for preload, Chrome: imagesrcset and imagesizes on link rel=preload • width/height on elements for • and more! 2020-2022

Fidelity vs. File-size Lossy vs. lossless Video by cottonbro from Pexels COMPRESSION

No content

ORIGINAL PNG - 700KB

Original - 700KB Squoosh.app

MOZJPEG - Q75 - 53KB

WEBP - Q75 - 50.6KB

AVIF - Q30 - 28.8KB

AVIF - Q38 - 46.9KB

AVIF - LOSSLESS - 610KB

WEBP - LOSSLESS - 535KB

Avid (4:4:4) @ 19.8KB THE 20KB CHALLENGE

JPEG - Q18 - 20KB 4:4:4

No content

8x8 BLOCKS HALOS DETAIL LOSS https://goo.gle/2Wsr16A JPEG DISTORTIONS

WEBP - Q24 - 20KB NO LOSSY 4:4:4

AVIF - Q25 - 19.8KB 4:4:4

Modern image formats • AVIF is a solid fi rst choice if lossy, low- fi delity compression is acceptable and saving bandwidth is the number one priority. Assuming encode/decode speeds meet your needs. • WebP is more widely supported and may be used for rendering regular images where advanced features like wide color gamut or text overlays are not required. • AVIF may not be able to compress non-photographic images as well as PNG. Compression savings from WebP may be lower than JPEG for high- fi delity lossy compression. • If both AVIF and WebP are not viable options, consider evaluating MozJPEG (optimize JPEG images), OxiPNG (non-photographic images), or JPEG 2000 (lossy or lossless photographic images) • Progressive enhancement via lets the browser choose the fi rst supported format in the order of preference. This implementation is considerably simpli fi ed when using image CDN’s where the Accept Header and content negotiation (e.g. auto-format and quality) can serve the best image.

SPEED Encode vs. Decode Static vs. Dynamic Loading

The modern element Lazy-load Async decode SET DIMENSIONS TO AVOID LAYOUT SHIFTS Placeholder

50% of the brain is involved in visual processing Time to get the sense of a visual scene 0.01s How much we recall of what we see or do vs. 20% of what we read. 80% Increase in willingness to read due to color visuals. 80% Neuroscience How much audiences more likely to be persuaded by a talk with visuals vs. a purely verbal one 17% To process a symbol and 100ms to attach meaning to it. 150ms Sources: goo.gle/imgsci

No content

There are many ways to load an image. https://goo.gle/3eLzPux

MEASURE & OPTIMIZE

Images impact Web Vitals in a number of ways

Speed Metrics Changelog https://goo.gle/csmlog

Optimal image serving for Web Vitals • For a fast Largest Contentful Paint • Request your key hero image early • Use , srcset + e ff i cient modern image formats • Avoid wasting pixels (compress, don't serve overly high DPR images) • Lazy-load o ff screen images (reduce network contention for key resources) • For a low Cumulative Layout Shift • Set dimensions (width, height) on your images • Use CSS aspect-ratio or aspect-ratio boxes to reserve space otherwise • For a low impact First Input Delay • Avoid images causing network contention with other critical resources like CSS and JS.

Largest Contentful Paint Element Lighthouse Audit function getLCPDebugTarget(entries) { const lastEntry = entries[entries.length - 1]; return lastEntry.element; } LCP API (RUM) web.dev/debug-web-vitals-in-the- fi eld/ User devices have di ff erent screen resolutions, which results in di ff erent page layouts and thus di ff erent elements being visible within the viewport. Content may be personalized for the current user, so the LCP candidate element could vary wildly from user to user. 🔬 ⛰

https://twitter.com/TimVereecke/status/1372210487191085063 https://twitter.com/OliverJAsh/status/1441095716227543042

Images & Developer tooling Identify images that could be served in a more modern format

Serving Modern Image Formats

Picking an image format • Compression • Speed • Limitations • Animation • Features • Tooling

JPEG 80 vs AVIF 70 vs WEBP 85 54% size reduction vs JPEG 26% size reduction vs JPEG https://www.industrialempathy.com/posts/avif-webp-quality-settings/

When lossy compression is good enough, AVIF can deliver significantly better results than the existing web codecs, including WebP.

Picking an image format Speed & Features

Image Optimization Tooling • Squoosh • Imagemin • ImageMagick • Sharp / Jimp • Thumbor • ImageOptim • XnConvert • Image CDNs

No content

No content

No content

No content

Images & Developer tooling Image format debugging

Identify images that can be better sized Properly size images

Images without dimensions

Layout Shifts from images without dimensions Lighthouse audit function getCLSDebugTarget(entries) { const largestShift = entries.reduce(( return a && a.value > b.value ? a : }); if (largestShift && largestShift.sour const largestSource = largestShift. return a.node && a.previousRect.w b.previousRect.width * b.prev }); if (largestSource) { return largestSource.node; } } PerformanceObserver

Lots of factors determine the best image size • Many factors determine the best image size • Screen size and resolution • Viewport • Layout and responsive breakpoints • Device pixel ratio (physical resolution) • Art direction Can one size fi t all?

Responsive Images: srcset and sizes

by width and density Responsive images

Device Pixel Ratio (DPR) / Pixel Density Capping Responsive images

Image Aspect Ratio Reserve space for responsive and lazy-loaded images Padding-top hack Mapped aspect-ratio Explicit aspect-ratio

Enhanced image preview in DevTools File size, rendered size, rendered aspect ratio, intrinsic size, intrinsic aspect ratio

Image lazy-loading Defer o ff screen images

Image lazy-loading Eager image fetching

Lazy-loading responsive images Lazy-loading images

Image lazy-loading Lazy image fetching

Only lazy-load offscreen images

Image discovered late - loaded via JS LCP image preloaded

Preload Largest Contentful Paint image Lighthouse audit

Request your image early

https://twitter.com/patmeenan/status/1443219921522073601

Blurry placeholder images Approximate the fi nal image before it loads in

Placeholder using background-image Placeholder Final image

Video by KoolShooters from Pexels IMAGES ARE SO COMPLICATED

Next.js Image Component • Automatic srcsets for images • Enforced sizing to prevent layout shift • Automatic lazy-loading of o ff screen assets • Framework-speci fi c asset prioritization optimizations • Flexible layout options • Customizable loader support for any image server or CDN Features and bene fi ts Up to 60% faster LCP and up to 100% reduced CLS

Next.js Image Component Features and bene fi ts • Automatic srcsets for images • Enforced sizing to prevent layout shift • Automatic lazy-loading of o ff screen assets • Framework-speci fi c asset prioritization optimizations • Flexible layout options • Customizable loader support for any image server or CDN Up to 60% faster LCP and up to 100% reduced CLS

Images in Next.js without Next Image import Head from 'next/head'; export default function Index() { return (

Next.js Image Component

Images in Next.js with Next Image import Head from 'next/head'; import Image from 'next/image'; export default function Index() { return (

Next.js Image Component

Next.js Image Component

THE BEST OPTIMIZATIONS IMPROVE USER-EXPERIENCE AND BUSINESS OUTCOMES.

31% improvement in LCP increased sales by 8% Optimizations to improve LCP included resizing their hero image, optimizing SVGs and using media queries to limit loading o ff screen images. Their optimizations to LCP included a 2.5s saving from switching their fi rst large image from being behind JavaScript (client-side hydration) to being directly in the main HTML document. 70% improvement in LCP correlated with a 76% reduction in load abandonment They determined shifts were caused after their hero images were loaded and snapped in for the fi rst view. They used Aspect Ratio Boxes to reserve space before their image was loaded. Optimizations to CLS helped increase News page views per session by 15% French Fashion house Chloè used Link Preload to preload their 1x and 2x Hero images, which were previously bottlenecked by a render-blocking script. Preloading 1x and 2x Hero images bottlenecked by render- blocking script improved LCP by 500ms (based on CrUX)

AUTOMATE BEST PRACTICES Video by Mikhail Nilov from Pexels

Thank you! @addyosmani