Our homepage opens with a moving ribbon of colour behind the headline. On a phone, the browser used to download a 2160-pixel video for that background. Before changing the artwork, we asked a more useful question. What does the visitor need for the very first frame, and what can arrive once that frame is on screen?
Our first English mobile Lighthouse test recorded the first visible content at 2.9 seconds and the largest visible element at 6.2 seconds. Those measurements describe different moments. The second figure helped us identify the video as the place to investigate. It could not tell us how many visitors left before seeing it.
If you want to run the same check, take a screen recording of a first visit on a phone. Play it back slowly and note three moments: when you can tell whose site it is, when you can read what they offer, and when the video starts moving. On a portfolio site, seeing the headline and the work can matter more than starting the loop a fraction earlier. A single performance score cannot show that sequence.
See the video on our homepage.
We exported a still image from the actual opening frame and told the browser to fetch it early. A phone now gets a 720-pixel version of the same six-second loop, about 229 KB instead of 1.4 MB. A larger screen can still use the original. Matching the crop and colours matters: an unrelated placeholder will visibly jump when playback begins.
Choose the still with the headline in place. We used the first frame because it makes the handover to this particular loop hard to spot. With other footage, that frame might be black, blurry or mid-transition. In that case, trim the beginning of the clip and export the still from the new starting point. Check the result on the narrowest screen you support. A bright area that sits harmlessly beside the text on desktop can move directly behind it when the video is cropped on a phone.
A smaller video solved only part of the delay. Our page builder kept the hero hidden until JavaScript started its entrance animation. Moving that slide and fade into CSS let the browser begin the entrance earlier. We also found that the Latvian homepage requested fonts from the English domain before loading the copies it actually used.
There is a simple way to tell whether another round of compression is worth it. Look at the browser’s network timeline for a fresh visit. If the first frame finishes downloading early but stays invisible, try disabling just the entrance effect in a test copy. If the artwork appears much sooner, the next change belongs in the animation or layout. If the frame itself arrives late, investigate when it was requested and what was downloading ahead of it.
In our November 2026 mobile lab checks, the English hero appeared in 3.1–3.5 seconds and the Latvian one in 4.0–4.1 seconds after the changes. These are repeatable test conditions, not measurements of every customer visit. For your own site, open each language on a phone, watch the transition into playback, and check that the headline and primary link are usable while the rest of the page loads.
How Google measures Largest Contentful Paint
Add a mobile acceptance test to your website brief.
Only until playback starts. Test what happens if autoplay is unavailable too. On that visit the still may be the only background the person sees, so the headline and button need to work against it for more than a split second.
Changing the displayed width can leave the phone downloading the entire desktop file. Open the network panel on a narrow viewport and check the actual filename and transferred bytes. That is how we verified that phones received the smaller loop.
Use it to compare two versions under the same conditions. For sign-off, also try a real phone: read the headline, tap the main link and watch the loop restart. If the site has enough real-user data, check how visitors’ loading times compare with your lab result.