Google is leveraging machine learning to make Duo calls sound better
What you need to know
- Google is using machine learning to deal with packet loss of audio data on its video calling app.
- The feature's already available on Pixel 4 devices via the December feature drop.
- Other as-yet-unnamed phones are also slated to receive the same tech in the future.
Google's Duo is already "the best video calling service that you're not using" — even Samsung, it seems, has seen light — but the company's not just resting on its laurels with a steady stream of new features rolling out every so often.
The latest innovation from Google in this regard has to also be one of the most fascinating, as the company is using machine learning algorithms to deal with packet loss (when data is lost as it travels across the internet) by estimating what the lost audio may have been (via The Verge).
While such technologies already exist, Google's implementation, the brainchild of its DeepMind subsidiary, fares far better than its competition and can replicate larger amount of missing audio, the company claims. It's also been trained to work with 48 different languages.
The company notes that nearly 99% of all Duo calls experience at least some packet loss, with 10% of calls losing as much as 8% of audio data. This technology can help improve the audio quality of Duo calls, especially in developing countries, where internet connections are less stable. Moreover, all the processing is done on-device, which means your privacy is protected and the feature respects Duo's end-to-end encryption.
Google says its already enabled the tech for Duo on the Pixel 4 with the December feature drop, and is also planning on expanding its availability to other phones. It did not, unfortunately, provide the names of these phones.
Be an expert in 5 minutes
Get the latest news from Android Central, your trusted companion in the world of Android