Lazy load images and <iframe> elements

Images and <iframe> elements often consume more bandwidth than other types of resources. In the case of <iframe> elements, a fair amount of extra processing time can be involved in loading and rendering the pages within them.

In the case of lazy loading images, deferring the loading of images that are outside of the initial viewport can be helpful in reducing bandwidth contention for more critical resources within the initial viewport. This can improve a page's Largest Contentful Paint (LCP) in some cases where network connections are poor, and that reallocated bandwidth can help LCP candidates load and paint faster.

Where <iframe> elements are concerned, a page's Interaction to Next Paint (INP) can be improved during startup by lazy loading them. This is because an <iframe> is a completely separate HTML document with its own subresources. While <iframe> elements can be run in a separate process, it's not uncommon for them to share a process with other threads, which can create conditions where pages become less responsive to user input.

Thus, deferring the loading of off-screen images and <iframe> elements is a technique worth pursuing, and requires fairly low effort for a reasonably good return in terms of performance. This module covers explains to lazy load these two types of elements for a faster and better user experience during the page's critical startup period.

Lazy load images with the loading attribute

The loading attribute can be added to <img> elements to tell browsers how they should be loaded:

  • "eager" informs the browser that the image should be loaded immediately, even if it's outside the initial viewport. This is also the default value for the loading attribute.
  • "lazy" defers loading an image until it's within a set distance from the visible viewport. This distance varies by browser, but is often set to be large enough that the image loads by the time the user scrolls to it.

It's also worth noting that if you're using the <picture> element, the loading attribute should still be applied to its child <img> element, not the <picture> element itself. This is because the <picture> element is a container that contains additional <source> elements pointing to different image candidates, and the candidate that the browser chooses is applied directly to its child <img> element.

Don't lazy load images that are in the initial viewport

You should only add loading="lazy" attribute to <img> elements that are positioned outside the initial viewport. However, it can be complex to know the precise position of an element relative within the viewport before the page is rendered. Different viewport sizes, aspect ratios, and devices have to be considered.

For example, a desktop viewport can be quite different from a viewport on a mobile phone as it renders more vertical space which may be able to fit images in the initial viewport that wouldn't appear in the initial viewport of a physically smaller device. Tablets used in portrait orientation also display a considerable amount of vertical space, perhaps even more than some desktop devices.

However, there are some cases in which it's fairly clear that you should avoid applying loading="lazy". For example, you should definitely omit the loading="lazy" attribute from <img> elements in cases of hero images, or other image use cases where <img> elements are likely to appear above the fold, or near the top of the layout on any device. This is even more important for images that are likely to be LCP candidates.

Images that are lazy loaded need to wait for the browser to finish layout in order to know if the image's final position is within the viewport. This means that if an <img> element in the visible viewport has a loading="lazy" attribute, it's only requested after all CSS is downloaded, parsed, and applied to the page—as opposed to being fetched as soon as it is discovered by the preload scanner in the raw markup.

Since the loading attribute on the <img> element is supported on all major browsers there is no need to use JavaScript to lazy load images, as adding extra JavaScript to a page to provide capabilities the browser already provides affects other aspects of page performance, such as INP.

Image lazy loading demo

Lazy load <iframe> elements

Lazy loading <iframe> elements until they are visible in the viewport can save significant data and improve the loading of critical resources that are required for the top-level page to load. Additionally, because <iframe> elements are essentially entire HTML documents loaded within a top-level document, they can include a significant number of subresources—particularly JavaScript—which can affect a page's INP considerably if the tasks within those frames requires significant processing time.

Third-party embeds are a common use case for <iframe> elements. For example, embedded video players or social media posts commonly use <iframe> elements, and they often require a significant number of subresources which can also result in bandwidth contention for the top-level page's resources. As an example, lazy loading a YouTube video's embed saves more than 500 KiB during the initial page load, while lazy loading the Facebook Like button plugin saves more than 200 KiB—most of which is JavaScript.

Either way, whenever you have an <iframe> below the fold on a page, you should strongly consider lazy loading it if it's not critical to load it up front, as doing so can significantly improve the user experience.

The loading attribute for <iframe> elements

The loading attribute on <iframe> elements is also supported in all major browsers. The values for the loading attribute and their behaviors are the same as with <img> elements that use the loading attribute:

  • "eager" is the default value. It informs the browser to load the <iframe> element's HTML and its subresources immediately.
  • "lazy" defers loading the <iframe> element's HTML and its subresources until it is within a predefined distance from the viewport.

Lazy loading iframes demo

Facades

Instead of loading an embed immediately during page load, you can load it on demand in response to a user interaction. This can be done by showing an image or another appropriate HTML element until the user interacts with it. Once the user interacts with the element, you can replace it with the third-party embed. This technique is known as a facade.

A common use case for facades is video embeds from third-party services where the embed may involve loading many additional and potentially expensive subresources—such as JavaScript—in addition to the video content itself. In such a case—unless there's a legitimate need for a video to autoplay—video embeds require the user to interact with them before playback by clicking the play button.

This is a prime opportunity to show a static image that is visually similar to the video embed and save significant bandwidth in the process. Once the user clicks on the image, it's then replaced by the actual <iframe> embed, which triggers the third-party <iframe> element's HTML and its subresources to begin downloading.

In addition to improving initial page load, another key upside is that if the user never plays the video, the resources required to deliver it are never downloaded. This is a good pattern, as it ensures the user only downloads what they actually want to, without making possibly faulty assumptions about the user's needs.

Chat widgets are is another excellent use case for the facade technique. Most chat widgets download significant amounts of JavaScript that can negatively affect page load and responsiveness to user input. As with loading anything up front, the cost is incurred at load time, but in the case of a chat widget, not every user never intends to interact with it.

With a facade on the other hand, it's possible to replace the third-party "Start Chat" button with a fake button. Once the user meaningfully interacts with it—such as holding a pointer over it for a reasonable period of time, or with a click—the actual, functional chat widget is slotted into place when the user needs it.

While it's certainly possible to build your own facades, there are open source options available for more popular third parties, such as lite-youtube-embed for YouTube videos, lite-vimeo-embed for Vimeo videos, and React Live Chat Loader for chat widgets.

JavaScript lazy loading libraries

If you need to lazy load <video> elements, <video> element poster images, images loaded by the CSS background-image property, or other unsupported elements, you can do so with a JavaScript-based lazy loading solution, such as lazysizes or yall.js, as lazy loading these types of resources is not a browser-level feature.

In particular, autoplaying and looping <video> elements without an audio track are a much more efficient alternative than using animated GIFs, which can often be several times larger than a video resource of equivalent visual quality. Even so, these videos can still be significant in terms of bandwidth, so lazy loading them is an additional optimization that can go a long way to reducing wasted bandwidth.

Most of these libraries work using the Intersection Observer API—and additionally the Mutation Observer API if a page's HTML changes after the initial load—to recognize when an element enters the user's viewport. If the image is visible—or approaching the viewport—then the JavaScript library replaces the non-standard attribute, (often data-src or a similar attribute), with the correct attribute, such as src.

Say you have a video that replaces an animated GIF, but you want to lazy load it with a JavaScript solution. This is possible with yall.js with the following markup pattern:

<!-- The autoplay, loop, muted, and playsinline attributes are to
     ensure the video can autoplay without user intervention. -->
<video class="lazy" autoplay loop muted playsinline width="320" height="480">
  <source data-src="video.webm" type="video/webm">
  <source data-src="video.mp4" type="video/mp4">
</video>

By default, yall.js observes all qualifying HTML elements with a class of "lazy". Once yall.js is loaded and executed on the page, the video doesn't load until the user scrolls it into the viewport. At that point, the data-src attributes on the <video> element's child <source> elements are swapped out to src attributes, which sends a request to download the video and automatically begin playing it.

Test your knowledge

Which is the default value for the loading attribute for both <img> and <iframe> elements?

"eager"
Correct!
"lazy"
Try again.

When are JavaScript-based lazy loading solutions reasonable to use?

For any resource that can be lazy loaded.
Try again.
For resources in which the loading attribute isn't supported, such as in the case of autoplaying videos intended to replace animated images, or to lazy load a <video> element's poster image.
Correct!

When is a facade a useful technique?

For any third-party embed that consumes significant data, regardless of the user's needs.
Try again.
For any third-party embed where the resources required to load are not only substantial, but there is a decent probability that not all users may interact with them.
Correct!

Up next: Prefetching and prerendering

Now that you have a handle on lazy loading images and <iframe> elements, you're in a good position to ensure that pages can load more quickly while respecting the needs of your users. However, there are cases in which speculative loading of resources can be desirable. In the next module, learn about prefetching and prerendering, and how these techniques—when used carefully—can substantially speed up navigations to subsequent pages by loading them ahead of time.