Boris Smus

interaction engineering

Remote controls for web media

When the world wide web was first conceived, it was as a collection of interlinked textual documents. Today's web is full of rich media. YouTube and other video sites alone consume an enormous 53% of all internet traffic. Web denizens often have an open audio player in one of their tabs. Web-based photo sharing services such as Flickr are the most common way of enjoying photos on our computers. The remote control, foundations of which are attributed to everyone's favorite inventor Nikola Tesla in patent US613809, has been the preferred way of controlling media for over half a century.

Yet the only way we can control all of this web media is via the on-screen user interfaces that the websites provide. The web has no remote control, and this is a big usability problem. Many use the desktop versions of streaming services like Spotify and Rdio rather than their web player, exclusively because of mac media key support. For scenarios where you're far from the screen, like showing friends a slideshow of photos on a TV, the lack of remote controllability is a non-starter.

This post is a concrete proposal for what a remote controls for the web should be like. To get a sense for how it might feel, try a rough prototype.

Ways of controlling media: dedicated keyboard buttons,headphone
remotes, hardware remote controls, second-screen remote controls,
camera-based gestures, voice commands

Related attempts to solve this problem

Many platforms, especially Android, Mac and iOS, do a pretty good job of supporting some of the inputs from the above image. The web, one of the most common platforms for consuming media, supports none of them. The only exception, of course, is the mouse and keyboard, but only when the player tab is in the foreground.

On the web, there have been a number of proposals and half-solutions to address this problem. Back in 2011, I shared KeySocket, a Menu Bar app for OS X that handles media keys on the mac keyboard and sends them to a companion Chrome extension that injects content scripts into web-based media players. A similar project, built support for media keys as an NPAPI plugin (a now deprecated technology). The Flutter app takes a similar approach (native app and companion extension), but enables webcam-based gestures for controlling media.

Recently, I contributed the Mac implementation to the new global keyboard shortcuts API for Chrome Apps and Extensions. This API lets developers bind to any global shortcut, including media keys. This is a good start since it brings the media key handling infrastructure into Chrome, but does not address the problem for the web in general.

Starting with a good user experience

Since we have a blank slate when it comes to controlling media on the web, how should media controls behave? Let's start with some sub-optimal behaviors. Here's one: all media events to get routed to all open tabs capable of handling them. Imagine the case with many YouTube tabs open, and the ensuing cacophony! Another bad approach is to route commands only to the foreground tab, since a very common case for needing media controls occurs when music is playing in the background.

This rough prototype illustrates what I think is a pretty good experience. It follows a focus-based model inspired by mobile operating systems like iOS and Android. However, the web is messier than the app model and edge cases like multiple sources of media playing simultaneously (eg. music player and YouTube video) are likely to happen, so we need to be careful.

Here is what happens when a user issues a play/pause command. I'll define the bold terms in a second.

  1. If any media is playing in a background tab, it should pause.
  2. Otherwise, if the foreground tab supports media events, it should receive the media control and be pushed to the media focus stack.
  3. Otherwise, if the media focus stack is non-empty, the event should be routed to the tab at the top of the stack.
  4. Otherwise (if the stack is empty), find the first open tab supporting media events, relay the event to that page and push it on the media focus stack.
  5. If there are no open tabs supporting media events, do nothing. Optionally alert the user with a non-modal notification (eg. audible chime).

When a next or previous control is issued, the command should be routed to the tab with media focus. If there are no tabs with media focus and none capable of media control, we drop the event.

If a tab closes, remove it from the media focus stack and ensure that media focus is granted to the tab at the top of the stack.

To clarify the description above, here are a few terms:

  • Foreground: the active tab of the foreground browser window.
  • Background: every tab that is not in the foreground.
  • Media events: a new event type that a page can listen to, indicating how to interpret media controls (see the next section).
  • Playing tab: a tab that is currently playing audio or video content.
  • Media focused tab: the tab which is the default receiver of media control events.
  • Media focus stack: a stack of tabs where the top-most tab is the one that currently has media focus. If that tab is popped off the stack, the next one gets media focus.

The dry description above and the prototype should give a sense of what tab should handle basic media controls, regardless of their origin: keyboard, remote control hardware, gesture, etc.

Now, when a command comes in, how does the page know how to interpret it? That's up to the web developer, and is done through media events, described in the next section.

Enabling media controls using media events

A fundamental missing piece so far is a way for a web page to indicate that it can receive media controls, and a way for it to specify how it wants to handle them. The solution to this is to create a new type of event, the media event, which are defined on a page-level, bound to the window object. This suggestion is not new, and first (as far as I can tell) came up in this blog post by Paul Rouget. Here's how media events work:

// Subscribing to media events.
window.addEventListener('media', function(e) {
  if ( == e.MEDIA_PLAY) {;
  } else if ( == e.MEDIA_PAUSE) {
  } else if ( == e.MEDIA_NEXT_TRACK) {;
  } else if ( == e.MEDIA_PREVIOUS_TRACK) {

This code tells the browser that this page can accept media controls, and what this page should do when a particular media control is received.

Determining user-initiated media playback change

Another missing piece in the narrative so far is how to populate the focus stack. So far, we know that a closed tab should be popped from the stack, and that play/pause sometimes causes a tab to be pushed onto the stack. But this is not enough, since the user can still interact with media using the UI of the player. For example, if I start listening to music through a Spotify tab, and then switch tabs, media commands should obviously go to the Spotify tab, despite me never having issued any media controls.

One option is to, as the user navigates between tabs, push any tab with supporting media events onto the stack. This approach fails for the case where you are listening to music in the background, and then change tabs, passing an open YouTube video on the way. In this case, that YouTube video would become focused and there would be no way to control the music (until you close the YouTube tab). What we actually need is to be able to track when the user interacts with the page using the media player UI, in order to then push page to the media focus stack.

A browser can distinguish user-initiated events (like clicks and keyboard presses) from programatic ones (like a timer firing, or a page loading). iOS does this to prevent annoying pages from autoplaying music (remember the 90s?). Using the same idea, browsers can also track when a media player's playback state changes due to user input.

Even so, there may be special cases that aren't perfect. For example, imagine a music app with media controls and a video ad on the side. If the user then clicks the video ad, it doesn't mean that the page should now have media focus. There are other tricky cases such as a page full of videos. Here, if a user starts playing a particular video, and then wants to stop it using media controls, the expectation is that the same video pauses. If the web developer does not handle this case gracefully, another video may start playing concurrently.

Response at FOMS was positive

Pitching this idea at FOMS 2013 a few months ago, folks seemed receptive. There was an understanding that a lack of media controls on the web is a genuine problem. Additionally, I got good feedback on the solution, which helped to iterate and get to this stage. This is all very encouraging, and I've written this post to keep the discussion alive and keep the momentum going. To make remote controls for the web a reality, we need is a critical mass of interested browser implementers.

As always, thanks for reading, and let me know if you have thoughts or suggestions on this topic, especially if you make browsers for a living and want to help standardize!