Unless otherwise noted, changes described below apply to the newest Chrome beta channel release for Android, Chrome OS, Linux, macOS, and Windows. Learn more about the features listed here through the provided links or from the list on ChromeStatus.com. Chrome 94 is beta as of August 26, 2021.
Existing media APIs (HTMLMediaElement, Media Source Extensions, WebAudio, MediaRecorder, and WebRTC) are high-level and narrowly-focused. A low-level codec API would better support emerging applications, such as latency-sensitive game streaming, client-side effects or transcoding, and polyfillable media container support, without the increased network and CPU cost of JavaScript or WebAssembly codec implementations.
HTMLMediaElement
WebAudio
MediaRecorder
The WebCodecs API eliminates these deficiencies by giving programmers a way to use media components that are already present in the browser. Specifically:
This feature has also completed its origin trial in Chrome 93 and is now available by default. For more information, see Video processing with WebCodecs.
The WebGPU API is the successor to the WebGL and WebGL2 graphics APIs for the Web. It provides modern features such as "GPU compute" as well as lower overhead access to GPU hardware and better, more predictable performance. This is an improvement over the existing WebGL interfaces, which were designed for drawing images but could only be repurposed for other kinds of computations with great effort.
WebGPU exposes modern computer graphics capabilities, specifically Direct3D 12, Metal, and Vulkan, for performing rendering and computation operations on a graphics processing unit (GPU). Advantages of WebGPU over earlier technologies include:
This feature is starting an origin trial in Chrome 94 with the hope of shipping in Chrome 99. For more information, see Access modern GPU features with WebGPU.
It's difficult to build web apps that are responsive to user interaction and that remain responsive over time. Scripts are one of the primary culprits hurting responsiveness. Consider a "search-as-you-type" feature: an app with this capability needs to keep up with the user's typing at the same time that it is fetching and displaying results. This doesn't take into account anything happening on the page such as animation, which must be rendered smoothly.
The problem is usually tackled by chunking and scheduling main thread work, specifically executing work asynchronously at appropriate times. This approach has its own problems, including the fact that whatever priority the developer sets, it's still competing for time on the main thread, which doesn't recognize the developer's prioritization, and is also responsible for browser tasks such as fetch() operations and garbage collection.
fetch()
The scheduler.postTask() method fixes these scheduling dilemmas by letting developers schedule tasks (JavaScript callbacks) with an OS browser scheduler at three levels of priority: user-blocking, user-visible, and background. It also exposes a TaskController interface, which can dynamically cancel tasks and change their priority.
scheduler.postTask()
TaskController
This feature completed its origin trial in Chrome 93 and is now available by default in Chrome. For a list of other new and completed origin trials, see the Origin Trials section below.
In addition to the items above, this version of Chrome introduces the origin trials described below. Origin trials allow you to try new features and give feedback on usability, practicality, and effectiveness to the web standards community. To register for any of the origin trials currently supported in Chrome, including the ones described below, visit the Chrome Origin Trials dashboard. To learn more about origin trials in Chrome, visit the Origin Trials Guide for Web Developers. Microsoft Edge runs its own origin trials separate from Chrome. To learn more, see the Microsoft Edge Origin Trials Developer Console.
Chrome is testing a new HTTP status code: 103 Early Hints for preloading subresources earlier. When a 103 response includes <link rel=preload> or other link headers Chromium tries to preload (and/or preconnect, prefetch) specified resources before the final response is received. This gives web developers a way to optimize apps, sites, and pages.
<link rel=preload>
The following features, previously in a Chrome origin trial, are now enabled by default.
This update formalizes that the default color space for CanvasRenderingContext2D objects and ImageData objects is sRGB. This clarifies that the CanvasRenderingContext2D interface is fully color managed (that all inputs are converted to the canvas color space). These were previously conventions and not clearly specified. This updates makes the following changes:
CanvasRenderingContext2D
ImageData
Content displayed by CanvasRenderingContext2D is currently limited to the sRGB color space, which is less than the capabilities of modern displays and cameras. This feature allows creating a CanvasRenderingContext2D object that is in the Display P3 color space. This also clears up several points of ambiguity about the color behavior of CanvasRenderingContext2D.
The VirtualKeyboard interface has methods and properties to control when a virtual keyboard is shown or hidden. It also fires events with the size of the virtual keyboard when it occludes content in the page. The virtual keyboard is the on-screen keyboard used for input in scenarios where a hardware keyboard may not be available.
VirtualKeyboard
Unlike a hardware keyboard, a virtual keyboard can adapt its shape to optimize for the expected input. Developers have control over the displayed shape of the virtual keyboard through the inputmode attribute, but have limited control over when the virtual keyboard is shown or hidden.
inputmode
The transform-style: preserve-3d and perspective properties now align with the spec. The preserve-3d property allows child elements to participate in the parent's 3D scene, and the perspective property applies a perspective transform to child elements. Before this change, Chromium applied both of these effects based on the containing block hierarchy rather than the DOM tree, and allowed them to extend through elements without transform-related properties on them.
Chrome now supports the keywords content, min-content, max-content, and fit-content as values for the flex-basis property and its flex shorthand. The content keyword makes flex base size use the default sizing rules as if flex-basis and preferred size property (width or height) are both auto, ignoring any specified width or height in the main axis dimension when flex-basis is auto. The other keywords are the same as usual and give more options for specifying the flex base size.
content
min-content
max-content
fit-content
flex-basis
flex
width
height
auto
In responsive layouts, when adding or removing display:flex to a container, you previously had to sometimes add/remove values for each individual item. content eliminates the need in some situations.
display:flex
The scrollbar-gutter property provides control over the presence of scrollbar gutters (the space reserved to display a scrollbar), allowing developers to prevent layout changes as content expands while avoiding unwanted visuals when scrolling isn't needed. Note that the presence of the scrollbars themselves is determined by the overflow property. The choice of classical or overlay scrollbars is up to the user agent. This property provides developers with more control over how their layouts interact with the scrollbars provided by the browser.
scrollbar-gutter
overflow
This API lets developers manipulate raw media carried by MediaStreamTracks such as the output of cameras, microphones, screen captures or the decoder part of a codec and the input to the decoder part of a codec. It uses WebCodecs interfaces to represent raw media frames and exposes them using streams, similar to the way the WebRTC Insertable Streams exposes encoded data from RTCPeerConnections. Example use cases include funny hats and real-time object identification and annotation.
MediaStreamTracks
With the removal of Flash, there is no longer a need to return anything for navigator.plugins and navigator.mimeTypes. These APIs were used primarily for:
navigator.plugins
navigator.mimeTypes
Some sites use these APIs to probe for PDF viewer support. With this change, these arrays will return fixed lists containing a standard list of PDF viewer plugins.
Note that this is not the removal or change of any API, it is merely the return of fixed arrays for these two existing APIs.
This version of Chrome incorporates version 9.4 of the V8 JavaScript engine. It specifically includes the change listed below. You can find a complete list of recent features in the V8 release notes.
Chrome now supports a web-exposed sampling profiler for measuring client JavaScript execution time. Gathering JavaScript profiles from real users can help developers debug slow observed performance without invasive manual instrumentation.
This version of Chrome introduces the deprecations and removals listed below. Visit ChromeStatus.com for lists of current deprecations and previous removals.
WebSQL in third-party contexts is now deprecated. Removal is expected in Chrome 97. The Web SQL Database standard was first proposed in April 2009 and abandoned in November 2010. Gecko never implemented this feature and WebKit deprecated in in 2019. The W3C encourages Web Storage and Indexed Database for those needing alternatives.
Developers should expect that WebSQL itself will be deprecated and removed when usage is low enough.
Private network requests for subresources may now only be initiated from a secure context. Private network requests are those initiated from a public network, targeting a private network. Examples include internet to intranet requests and intranet loopbacks.
This is a first step towards fully implementing Private Network Access. Servers running inside local networks, or on a user's device, expose powerful capabilities to the web in ways that can be quite dangerous. Private Network Access proposes a set of changes to limit the impact of requests to these servers by ensuring that the servers are opting-into any communication with external entities.
For this opt-in to have any meaning, the servers need to be able to ensure that the client origin is authenticated. To that end, only secure contexts are empowered to make external requests.
Unless otherwise noted, changes described below apply to the newest Chrome beta channel release for Android, Android WebView, Chrome OS, Linux, macOS, and Windows. Learn more about the features listed here through the provided links or from the list on ChromeStatus.com. Chrome 93 is beta as of July 29, 2021.
This version of Chrome introduces the origin trials described below. Origin trials allow you to try new features and give feedback on usability, practicality, and effectiveness to the web standards community. To register for any of the origin trials currently supported in Chrome, including the ones described below, visit the Chrome Origin Trials dashboard. To learn more about origin trials in Chrome, visit the Origin Trials Guide for Web Developers. Microsoft Edge runs its own origin trials separate from Chrome. To learn more, see the Microsoft Edge Origin Trials Developer Console.
Cross-origin no-CORS requests can now be made to omit credentials (cookies, client certificates, etc.) using the credentialless keyword. Similarly to COEP: require-corp, it can enable cross-origin isolation.
credentialless
COEP: require-corp
Sites that wish to continue using SharedArrayBuffer must opt-in to cross-origin isolation. Today, COEP: require-corp exists, and is used to enable cross-origin isolation. It is functional and solid, but turns out to be difficult to deploy at scale, as it requires all subresources to explicitly opt-in. This is fine for some sites, but creates dependency problems for sites that gather content from users (Google Earth, social media generally, forums, etc.).
SharedArrayBuffer
The Multi-Screen Window Placement API allows you to place windows on any display connected to your machine, store that placement, and make a window full-screen on any display. With this API, a presentation app can show slides on one screen and speaker notes on another. An art or music creation app can place palettes on a second screen. And a restaurant can show a touchscreen menu on a kiosk and a separate window for employees. Incorporating developer feedback from the first origin trial, the API enters a second origin trial with an improved shape and ergonomics.
Window controls overlay extends an app's client area to cover the entire window, including the title bar, and the window control buttons (close, maximize/restore, minimize). The web app developer is responsible for drawing and input-handling for the entire window except for the window controls overlay. Developers can use this feature to make their installed desktop web apps look like O.S. apps. For more information, see Customize the window controls overlay of your PWA's title bar.
PWA as URL Handlers allows apps like music.example.com to register themselves as URL handlers for URLs that match patterns like https://music.example.com, https://*.music.example.com, or https://🎵.example.com, so that links from outside of the PWA, for example, from an instant messenger application or an email client, open in the installed PWA rather than in a browser tab.
music.example.com
https://music.example.com
https://*.music.example.com
https://🎵.example.com
Web Bundles provides a new approach to load a large number of resources efficiently using a format that allows multiple resources to be bundled. This feature addresses issues with previous approaches to resource bundling.
The output of some JavaScript bundlers does not interact well with the HTTP cache and configuring them can sometimes be difficult. Even with bundled JavaScript, execution needs to wait for all bytes to download. Ideally loading multiple subresources should use streaming and parallelization, which is not possible with one JavaScript file. For JavaScript modules, execution still needs to wait for an entire resource tree to download because of deterministic execution.
WebXR applications can now retrieve data about planes present in a user's environment, which enables augmented reality applications to create more immersive experiences. Without this feature, developers would have to resort to running their own computer vision algorithms on data from getUserMedia() (available on navigator and MediaDevices) in order to detect planes present in the users' environment. Such solutions have thus far been unable to match the quality and accuracy of native augmented reality capabilities or support world scale.
getUserMedia()
navigator
MediaDevices
AbortSignal.abort() is a static method that allows creating a new AbortSignal object that is already aborted. It is similar in spirit to Promise.reject(), and provides improved developer ergonomics.
AbortSignal.abort()
AbortSignal
Promise.reject()
Web developers have found aborted AbortSignal objects to be useful for a variety of purposes. They signify to JavaScript APIs that no work should be done. Currently, creating an already-aborted AbortSignal object requires multiple lines of code. AbortSignal.abort() requires one: return AbortSignal.abort();
return AbortSignal.abort();
The flexbox and flex items now obey positional alignment keywords. Flexbox previously only obeyed center, flex-start, and flex-end. The additional alignment keywords (start, end, self-start, self-end, left, right) allow authors to more easily align the flex items in the face of varying writing modes and flex flows.
center
flex-start
flex-end
start
end
self-start
self-end
left
right
Without these additional keywords, developers need to change the keyword values whenever they change the writing mode, text direction, or flex reversal properties (flex-direction: row-reverse, flex-direction:column-reverse or align-content: wrap-reverse). The keywords implemented here let them set alignment once.
flex-direction: row-reverse
flex-direction:column-reverse
align-content: wrap-reverse
The Error() constructor supports a new options property called cause, which will be assigned to the error as a property. This allows errors to be chained without unnecessary and overelaborate formalities on wrapping the errors in conditions.
Error()
The meta element's "media" attribute will be honored for meta[name="theme-color"] so that web developers can adjust the theme color of their site based on a media query (dark and light modes for instance). The first one that matches will be picked.
meta[name="theme-color"]
The HTMLMediaElement.controlsList property now supports noplaybackrate, which allows websites to enabled or disable the playback speed control exposed by the browser. With browser vendors adding playback speed control to their media controls, developers should have a way to control the visibility of this new control. Try the new property on the noplaybackrate in HTMLMediaElement.controlsList Sample.
noplaybackrate
HTMLMediaElement.controlsList
The CSS user preference media feature prefers-color-scheme has a potentially significant impact on the amount of CSS that needs to be delivered by a page and on the experience the user is going to have when the page loads. The new Sec-CH-Prefers-Color-Scheme client hint header allows sites to obtain the user's preference optionally at request time, allowing servers to inline the right CSS and therefore avoid a flash of incorrect color theme.
prefers-color-scheme
Sec-CH-Prefers-Color-Scheme
This version of Chrome adds four new features and changes to the User-Agent client hints API.
Sec-CH-UA-Platform
Low-entropy hints are those that don't give away too much information, or give information that would be too easy to discover in other ways to realistically hide. In the context of client hints, this means that these hints are available in every request, whether or not the origin involved requested it or whether the frame involved is a first or third party context.
The WebOTP API will now be supported on desktop when both Chrome on Desktop and Android Chrome are logged in using the same Google account. The WebOPT API provides the ability to programmatically read a one-time code from specially-formatted SMS messages addressed to their origin, reducing user friction during sign-on. Previously, this was only available on mobile devices where SMS was supported.
This version of Chrome incorporates version 9.3 of the V8 JavaScript engine. It specifically includes the changes listed below. You can find a complete list of recent features in the V8 release notes.
Object.hasOwn, a new boolean property, provides an easier-to-use, static method version of Object.prototype.hasOwnProperty.
Object.hasOwn
Object.prototype.hasOwnProperty
Connections to HTTP, HTTPS or FTP servers on ports 989 and 990 now fail. These ports are used by the FTPS protocol, which has never been implemented in Chrome. However, FTPS servers can be attacked in a cross-protocol attack by malicious web pages using carefully-crafted HTTPS requests. This is a mitigation for the ALPACA attack.
Chrome has now removed support for the TLS_RSA_WITH_3DES_EDE_CBC_SHA cipher suite. TLS_RSA_WITH_3DES_EDE_CBC_SHA is a remnant of the SSL 2.0 and SSL 3.0 era. 3DES in transport layer security (TLS) is vulnerable to the Sweet32 attack. Being a CBC cipher suite, it is also vulnerable to the Lucky Thirteen attack. The first replacement AES cipher suites were defined for TLS in RFC3268, published around 19 years ago, and there have been several iterations since.
WebAssembly module sharing between cross-origin but same-site environments will be deprecated to allow agent clusters to be scoped to origins long term. This follows a WebAssembly specification change, which has an impact on the platform as well.
Unless otherwise noted, changes described below apply to the newest Chrome beta channel release for Android, Chrome OS, Linux, macOS, and Windows. Learn more about the features listed here through the provided links or from the list on ChromeStatus.com. Chrome 92 is beta as of June 3, 2021
Now that web apps are capable of reading and writing files, the next logical step is to let developers declare web apps as file handlers for files they create and process. The File Handling API allows you to do exactly this. For example, after a text editor PWA has registered itself as a file handler, you can right-click a .txt file in your operating system's file manager and instruct this PWA to (always or just once) open .txt files. This means PWAs are just a (double) click away from the file manager.
This improves the user experience for PWA use cases, making them more like OS apps than they are already. For example:
File Handling is starting an origin trial in 92 that is expected to run until around the end of August, 2021. For more information on this feature, see Let web applications be file handlers. For information about other origin trials in this release, see Origin Trials, below.
Shared Element Transitions allows a simple set of transitions in both single-page applications (SPAs) and multi-page applications (MPAs). This enhances the visual polish of pages with minimal effort by letting developers select from a set of user-agent provided transition effects. Without this, single-page app transitions are difficult since they require a careful coordination of animations and DOM manipulations to achieve the desired effect. Multi-page app transitions are, for the most part, not possible since each page can only control the contents of its own view. This origin trial is only for SPA use cases.
Most Android launchers now allow only three app shortcuts instead of the previously allowed four. A shortcut to the site settings was added to the application icon in the Android launcher, taking one of the available shortcut slots for the app. For more information, see Get things done quickly with app shortcuts.
Adds the size-adjust descriptor for @font-face allowing scaling of glyph sizes for a particular font face without affecting the CSS font-size and derived metrics such as em. CSS font-size can be seen as a scale factor for a box that the font draws in. Glyph sizes within that box vary between fonts, and size-adjust enables harmonising them across different fonts. That's why it reduces cumulative layout shift by matching up the fallback font and primary web font using this descriptor.
@font-face
font-size
Imperative slotting allows node-to-slot assignments without needing the slot attribute in markup. This enables dynamic slotting behavior based on input conditions and types. The feature originally shipped in Chrome 86; in this release some adjustments to the API have been made to ensure interoperability with other browsers.
This version of Chrome incorporates version 9.2 of the V8 JavaScript engine. It specifically includes the changes listed below. You can find a complete list of recent features in the V8 release notes.
A dayPeriod option (part of ECMA402 2021) has been added to the Intl.DateTimeFormat() method so the caller can format times such as "7 in the morning", "11 in the morning", "12 noon", "1 in the afternoon", "6 in the evening", "10 at night" (or in Chinese, "清晨7時", "上午11時", "中午12時", "下午1時" ,"下午6時" ,"晚上10時").
Intl.DateTimeFormat()
This enhances Intl.DateTimeFormat() to match what is already possible in C++ and Java by calling ICU and ICU4J. Without this feature, developers need to either format the quarter in the server or ship a set of day period patterns and hour to day period mappings from the server to client to perform such tasks.
Adds a new method named at(), to Array.prototype, String.prototype, and the TypedArray prototypes, that permit relative indexing with negative indices. For example: let arr = [1,2,3,4]; arr.at(-1); // Returns 4
at()
Array.prototype
String.prototype
TypedArray
let arr = [1,2,3,4]; arr.at(-1); // Returns 4
The ICU LocaleMatcher now implements the BestFitMatcher abstract operation to better match locale data.
SharedArrayBuffers on desktop platforms are now restricted to cross-origin isolated environments, matching the behavior recently shipped on Android and Firefox. A Cross-origin isolated page is considered a secure environment because it blocks loading cross-origin resources that are not opt-in and communicating with cross-origin windows. Only pages that opt-in to cross-origin isolation will be able to use SharedArrayBuffers. Learn more about the upcoming options at SharedArrayBuffer updates in Android Chrome 88 and Desktop Chrome 92.
SharedArrayBuffers
Adds actions to the Media Session API, specifically "togglemicrophone", "togglecamera", and "hangup". This enables developers of video conferencing websites to handle these actions from the browser interface. For example, if the user puts their video call into a picture-in-picture window, the browser could display buttons for mute/unmute, turn-on/turn-off camera, and hanging up. When the user clicks these, the website handles them through the Media Session API. For more information, see the section from our recent article, or try our demo.
"togglemicrophone"
"togglecamera"
"hangup"
Chrome now accounts for the tainted origin flag when computing whether a fetched resource passes the timing allow origin check. The timing allow origin check is used in resource timing to determine whether the page can receive detailed timing information about a resource used in the page. The tainted origin flag impacts this check in cases where there are multiple redirects that cross origins. In those cases, the header should be '*'. In other words, it can no longer be a specific origin.
If a resource performs two cross-origin crosses (via redirects), then the developer needs to use Timing-Allow-Origin: * for the checks to pass. For example if a page on origin A fetches a resource on origin B, which redirects to a resource on origin C, then the tainted origin flag is set and the final resource needs to have Timing-Allow-Origin: * in order to receive detailed timing information."
Timing-Allow-Origin: *
Web applications can now register themselves as handlers of custom URL protocols or schemes using their web app manifests. Operating system applications often register themselves as protocol handlers to increase discoverability and usage. Websites can already register to handle schemes via registerProtocolHandler(), it is desirable to have web apps be first-class citizens and be launched directly when a custom-scheme link is invoked.
The Web Bluetooth API can now filter based on manufacturer data such as vendor ID and product ID. Developers have been able to prompt users through a browser picker to select a nearby Bluetooth device that matches their advertised name and services. However it hasn't been possible to filter nearby Bluetooth devices based on advertised manufacturer specific data. Manufacturer data is specified through new properties on options.filters, which is passed to Bluetooth.requestDevice(). For more information, see Communicating with Bluetooth devices over JavaScript or try our demo.
Bluetooth.requestDevice()
This feature, which enabled web-based payment handlers to receive paymentrequest events with non-URL, but standardized payment method identifiers, such as "basic-card" or "tokenized-card", has been removed.
paymentrequest
"basic-card"
"tokenized-card"
