Google’s YouTube will use artificial intelligence to improve the visual clarity of many low-resolution videos on its platform when viewed on TV screens, the web and mobile devices.
YouTube will first implement the feature, known as upscaling, for videos that were originally uploaded at resolutions under 1080p, according to a blog post Oct 29 by senior product director Kurt Wilms. The company will support upscaling to even higher 4K quality "in the near future,” he added. A Google spokesperson confirmed the launch of upscaling will also include videos shown on YouTube’s web and mobile interfaces.
Some creators might take issue with the platform making adjustments to their content, but Wilms said that the original files will always be kept intact. YouTube members can opt out of the "super resolution” upscaling feature altogether if they don’t like how it makes their videos look, he said.
Super resolution videos will be clearly labeled, making it easy for viewers to see any video as it was originally presented, he added.
The introduction of upscaling should theoretically improve the YouTube experience for viewers streaming videos on large-screen TVs, wrote Wilms, who added that the TV screen is YouTube’s "fastest-growing surface.” But it could also make videos clearer on larger smartphones and foldables.
Many of the platform’s most popular creators have been publishing content at 4K resolution for several years now, but there’s an enormous trove of older clips that were uploaded before ultra high definition became the norm. Many top out at standard definition, known as 480p, or even lower.
YouTube is also greatly expanding the permitted file size of video thumbnails, which can now be 50 megabytes compared with the prior 2-megabyte limit, allowing for more detailed thumbnails on TV screens.
Beyond these changes, Wilms said the company has been testing "larger video uploads with select creators” that would lead to even higher fidelity as YouTube competes with Netflix Inc. and other streaming services for viewers’ time.
Over the summer, YouTube angered some creators when it quietly used machine learning to enhance the clarity of Shorts, its short-form videos meant to compete with content on TikTok and Instagram Reels. In that instance, YouTube clarified in a social media post that there was "no upscaling” at play. – Bloomberg
