How to accelerate your media playback by actively preloading resources.
Faster playback start means more people watching your video or listening to your audio.That's a known fact.In this article I'll explore techniques you can use to accelerate your audio and video playback by actively preloading resources depending on your use case.
Credits: copyright Blender Foundation |www.blender.org.
I'll describe three methods of preloading media files, starting with their pros and cons.
It's great... | But... | |
---|---|---|
Video preload attribute | Simple to use for a unique file hosted on a web server. | Browsers may completely ignore the attribute. |
Resource fetching starts when the HTML document has been completely loaded and parsed. | ||
Media Source Extensions (MSE) ignore thepreload attribute on media elements because the app is responsible for
providing media to MSE.
|
||
Link preload |
Forces the browser to make a request for a video resource without blocking
the document'sonload event.
|
HTTP Range requests are not compatible. |
Compatible with MSE and file segments. | Should be used only for small media files (<5 MB) when fetching full resources. | |
Manual buffering | Full control | Complex error handling is the website's responsibility. |
Video preload attribute
If the video source is a unique file hosted on a web server, you may want to
use the videopreload
attribute to provide a hint to the browser as tohow
much information or content to preload.This meansMedia Source Extensions
(MSE)is not compatible withpreload
.
Resource fetching will start only when the initial HTML document has been
completely loaded and parsed (e.g. theDOMContentLoaded
event has fired)
while the very differentload
event will be fired when resource
has actually been fetched.
Setting thepreload
attribute tometadata
indicates that the user is not
expected to need the video, but that fetching its metadata (dimensions, track
list, duration, and so on) is desirable. Note that starting inChrome
64,the default value forpreload
ismetadata
.(It wasauto
previously).
<video id= "video" preload= "metadata" src= "file.mp4" controls></video>
<script>
video.addEventListener('loadedmetadata', function() {
if (video.buffered.length === 0) return;
const bufferedSeconds = video.buffered.end(0) - video.buffered.start(0);
console.log(`${bufferedSeconds} seconds of video are ready to play.`);
});
</script>
Setting thepreload
attribute toauto
indicates that the browser may cache
enough data that complete playback is possible without requiring a stop for
further buffering.
<video id= "video" preload= "auto" src= "file.mp4" controls></video>
<script>
video.addEventListener('loadedmetadata', function() {
if (video.buffered.length === 0) return;
const bufferedSeconds = video.buffered.end(0) - video.buffered.start(0);
console.log(`${bufferedSeconds} seconds of video are ready to play.`);
});
</script>
There are some caveats though. As this is just a hint, the browser may completely
ignore thepreload
attribute. At the time of writing, here are some rules
applied in Chrome:
- WhenData Saveris enabled, Chrome forces the
preload
value tonone
. - In Android 4.3, Chrome forces the
preload
value tonone
due to anAndroid Bug. - On a cellular connection (2G, 3G, and 4G), Chrome forces the
preload
value tometadata
.
Tips
If your website contains many video resources on the same domain, I would
recommend you set thepreload
value tometadata
or define theposter
attribute and setpreload
tonone
.That way, you would avoid hitting
the maximum number of HTTP connections to the same domain (6 according to the
HTTP 1.1 spec) which can hang loading of resources. Note that this may also
improve page speed if videos aren't part of your core user experience.
Link preload
Ascoveredin otherarticles,link preloadis a declarative fetch that
allows you to force the browser to make a request for a resource without
blocking theload
event and while the page is downloading. Resources
loaded via<link rel= "preload" >
are stored locally in the browser, and are
effectively inert until they're explicitly referenced in the DOM, JavaScript,
or CSS.
Preload is different from prefetch in that it focuses on current navigation and fetches resources with priority based on their type (script, style, font, video, audio, etc.). It should be used to warm up the browser cache for current sessions.
Preload full video
Here's how to preload a full video on your website so that when your JavaScript asks to fetch video content, it is read from cache as the resource may have already been cached by the browser. If the preload request hasn't finished yet, a regular network fetch will happen.
<link rel= "preload" as= "video" href= "https://cdn.com/small-file.mp4" >
<video id= "video" controls></video>
<script>
// Later on, after some condition has been met, set video source to the
// preloaded video URL.
video.src = 'https://cdn.com/small-file.mp4';
video.play().then(() => {
// If preloaded video URL was already cached, playback started immediately.
});
</script>
Because the preloaded resource is going to be consumed by a video element in
the example, theas
preload link value isvideo
.If it were an audio
element, it would beas= "audio"
.
Preload the first segment
The example below shows how to preload the first segment of a video with<link
rel= "preload" >
and use it with Media Source Extensions. If you're not familiar
with the MSE JavaScript API, seeMSE basics.
For the sake of simplicity, let's assume the entire video has been split into
smaller files likefile_1.webm
,file_2.webm
,file_3.webm
,etc.
<link rel= "preload" as= "fetch" href= "https://cdn.com/file_1.webm" >
<video id= "video" controls></video>
<script>
const mediaSource = new MediaSource();
video.src = URL.createObjectURL(mediaSource);
mediaSource.addEventListener('sourceopen', sourceOpen, { once: true });
function sourceOpen() {
URL.revokeObjectURL(video.src);
const sourceBuffer = mediaSource.addSourceBuffer('video/webm; codecs= "vp09.00.10.08" ');
// If video is preloaded already, fetch will return immediately a response
// from the browser cache (memory cache). Otherwise, it will perform a
// regular network fetch.
fetch('https://cdn.com/file_1.webm')
.then(response => response.arrayBuffer())
.then(data => {
// Append the data into the new sourceBuffer.
sourceBuffer.appendBuffer(data);
// TODO: Fetch file_2.webm when user starts playing video.
})
.catch(error => {
// TODO: Show "Video is not available" message to user.
});
}
</script>
Support
You can detect support of variousas
types for<link rel=preload>
with the
snippets below:
function preloadFullVideoSupported() {
const link = document.createElement('link');
link.as = 'video';
return (link.as === 'video');
}
function preloadFirstSegmentSupported() {
const link = document.createElement('link');
link.as = 'fetch';
return (link.as === 'fetch');
}
Manual buffering
Before we dive into theCache APIand service workers, let's see
how to manually buffer a video with MSE. The example below assumes that your web
server supports HTTPRange
requests but this would be pretty similar with file
segments. Note that some middleware libraries such asGoogle's Shaka
Player,JW Player,andVideo.jsare
built to handle this for you.
<video id= "video" controls></video>
<script>
const mediaSource = new MediaSource();
video.src = URL.createObjectURL(mediaSource);
mediaSource.addEventListener('sourceopen', sourceOpen, { once: true });
function sourceOpen() {
URL.revokeObjectURL(video.src);
const sourceBuffer = mediaSource.addSourceBuffer('video/webm; codecs= "vp09.00.10.08" ');
// Fetch beginning of the video by setting the Range HTTP request header.
fetch('file.webm', { headers: { range: 'bytes=0-567139' } })
.then(response => response.arrayBuffer())
.then(data => {
sourceBuffer.appendBuffer(data);
sourceBuffer.addEventListener('updateend', updateEnd, { once: true });
});
}
function updateEnd() {
// Video is now ready to play!
const bufferedSeconds = video.buffered.end(0) - video.buffered.start(0);
console.log(`${bufferedSeconds} seconds of video are ready to play.`);
// Fetch the next segment of video when user starts playing the video.
video.addEventListener('playing', fetchNextSegment, { once: true });
}
function fetchNextSegment() {
fetch('file.webm', { headers: { range: 'bytes=567140-1196488' } })
.then(response => response.arrayBuffer())
.then(data => {
const sourceBuffer = mediaSource.sourceBuffers[0];
sourceBuffer.appendBuffer(data);
// TODO: Fetch further segment and append it.
});
}
</script>
Considerations
As you're now in control of the entire media buffering experience, I suggest you consider the device's battery level, the "Data-Saver Mode" user preference and network information when thinking about preloading.
Battery awareness
Take into account the battery level of users' devices before thinking about preloading a video. This will preserve battery life when the power level is low.
Disable preload or at least preload a lower resolution video when the device is running out of battery.
if ('getBattery' in navigator) {
navigator.getBattery()
.then(battery => {
// If battery is charging or battery level is high enough
if (battery.charging || battery.level > 0.15) {
// TODO: Preload the first segment of a video.
}
});
}
Detect "Data-Saver"
Use theSave-Data
client hint request header to deliver fast and light
applications to users who have opted-in to "data savings" mode in their
browser. By identifying this request header, your application can customize and
deliver an optimized user experience to cost- and performance-constrained
users.
SeeDelivering Fast and Light Applications with Save-Datato learn more.
Smart loading based on network information
You may want to checknavigator.connection.type
prior to preloading. When
it's set tocellular
,you could prevent preloading and advise users that
their mobile network operator might be charging for the bandwidth, and only start
automatic playback of previously cached content.
if ('connection' in navigator) {
if (navigator.connection.type == 'cellular') {
// TODO: Prompt user before preloading video
} else {
// TODO: Preload the first segment of a video.
}
}
Check out theNetwork Information sampleto learn how to react to network changes as well.
Pre-cache multiple first segments
Now what if I want to speculatively pre-load some media content without
knowing which piece of media the user will eventually pick? If the user is on a
web page that contains 10 videos, we probably have enough memory to fetch one
segment file from each but we should definitely not create 10 hidden<video>
elements and 10MediaSource
objects and start feeding that data.
The two-part example below shows you how to pre-cache multiple first segments of
video using the powerful and easy-to-useCache API.Note that something similar
can be achieved with IndexedDB as well. We're not using service workers yet as
the Cache API is also accessible from thewindow
object.
Fetch and cache
const videoFileUrls = [
'bat_video_file_1.webm',
'cow_video_file_1.webm',
'dog_video_file_1.webm',
'fox_video_file_1.webm',
];
// Let's create a video pre-cache and store all first segments of videos inside.
window.caches.open('video-pre-cache')
.then(cache => Promise.all(videoFileUrls.map(videoFileUrl => fetchAndCache(videoFileUrl, cache))));
function fetchAndCache(videoFileUrl, cache) {
// Check first if video is in the cache.
return cache.match(videoFileUrl)
.then(cacheResponse => {
// Let's return cached response if video is already in the cache.
if (cacheResponse) {
return cacheResponse;
}
// Otherwise, fetch the video from the network.
return fetch(videoFileUrl)
.then(networkResponse => {
// Add the response to the cache and return network response in parallel.
cache.put(videoFileUrl, networkResponse.clone());
return networkResponse;
});
});
}
Note that if I were to use HTTPRange
requests, I would have to manually recreate
aResponse
object as the Cache API doesn't supportRange
responsesyet.Be
mindful that callingnetworkResponse.arrayBuffer()
fetches the whole content
of the response at once into renderer memory, which is why you may want to use
small ranges.
For reference, I've modified part of the example above to save HTTP Range requests to the video precache.
...
return fetch(videoFileUrl, { headers: { range: 'bytes=0-567139' } })
.then(networkResponse => networkResponse.arrayBuffer())
.then(data => {
const response = new Response(data);
// Add the response to the cache and return network response in parallel.
cache.put(videoFileUrl, response.clone());
return response;
});
Play video
When a user clicks a play button, we'll fetch the first segment of video available in the Cache API so that playback starts immediately if available. Otherwise, we'll simply fetch it from the network. Keep in mind that browsers and users may decide to clear theCache.
As seen before, we use MSE to feed that first segment of video to the video element.
function onPlayButtonClick(videoFileUrl) {
video.load(); // Used to be able to play video later.
window.caches.open('video-pre-cache')
.then(cache => fetchAndCache(videoFileUrl, cache)) // Defined above.
.then(response => response.arrayBuffer())
.then(data => {
const mediaSource = new MediaSource();
video.src = URL.createObjectURL(mediaSource);
mediaSource.addEventListener('sourceopen', sourceOpen, { once: true });
function sourceOpen() {
URL.revokeObjectURL(video.src);
const sourceBuffer = mediaSource.addSourceBuffer('video/webm; codecs= "vp09.00.10.08" ');
sourceBuffer.appendBuffer(data);
video.play().then(() => {
// TODO: Fetch the rest of the video when user starts playing video.
});
}
});
}
Create Range responses with a service worker
Now what if you have fetched an entire video file and saved it in
the Cache API? When the browser sends an HTTPRange
request, you certainly don't
want to bring the entire video into renderer memory as the Cache API doesn't
supportRange
responsesyet.
So let me show how to intercept these requests and return a customizedRange
response from a service worker.
addEventListener('fetch', event => {
event.respondWith(loadFromCacheOrFetch(event.request));
});
function loadFromCacheOrFetch(request) {
// Search through all available caches for this request.
return caches.match(request)
.then(response => {
// Fetch from network if it's not already in the cache.
if (!response) {
return fetch(request);
// Note that we may want to add the response to the cache and return
// network response in parallel as well.
}
// Browser sends a HTTP Range request. Let's provide one reconstructed
// manually from the cache.
if (request.headers.has('range')) {
return response.blob()
.then(data => {
// Get start position from Range request header.
const pos = Number(/^bytes\=(\d+)\-/g.exec(request.headers.get('range'))[1]);
const options = {
status: 206,
statusText: 'Partial Content',
headers: response.headers
}
const slicedResponse = new Response(data.slice(pos), options);
slicedResponse.setHeaders('Content-Range': 'bytes ' + pos + '-' +
(data.size - 1) + '/' + data.size);
slicedResponse.setHeaders('X-From-Cache': 'true');
return slicedResponse;
});
}
return response;
}
}
It is important to note that I usedresponse.blob()
to recreate this sliced
response as this simply gives me a handle to the file while
response.arrayBuffer()
brings the entire file into renderer memory.
My customX-From-Cache
HTTP header can be used to know whether this request
came from the cache or from the network. It can be used by a player such as
ShakaPlayerto ignore the response time as an indicator of
network speed.
Have a look at the officialSample Media Appand in particular its
ranged-response.jsfile for a complete solution for how to handleRange
requests.