According to a Google AI blog post, a new audio codec, called Lyra, was developed by the team, with the aim of compressing speech at a lower bitrate. At 3 kbps, Lyra uses lower data than most other codecs. Google AI noted that codecs capable of operating at bitrates comparable to Lyra's suffer from "increased artifacts and result in a robotic voice".
Currently, the open source codec Opus, the most used codec for VOIP applications, with audio at 32kbps, can achieve transparent voice quality. While Opus can be used in more bandwidth-constrained environments, even at 6kbps, it is starting to demonstrate degraded audio quality, Google AI said. This makes the Opus codec less preferable than Lyra at 3 kbps.
CODECs improve the call experience
To hear Lyra and Opus working inside the Google Duo, if you care how they compare, check out the Google AI blog post. Google AI said it used a combination of existing codec technology and "advances in machine learning with models trained on thousands of hours of data" - including speakers in more than 70 languages from open source libraries - when working on Lyra. The team plans to continue developing the codec and hopes that it will be adopted by developers and applications beyond Google Duo.
Meanwhile, Google AI plans to release Lyra (on Google Duo) to boost audio calls over very low bandwidth connections. It is unknown when the new Lyra codec will be widely available to Duo users, although it will likely be a silent update in the background.
If you are a Google Duo user on Android or iOS with poor or unreliable internet connections, you will soon notice a drastic improvement in your Duo calls in terms of quality and stability. It is unclear when this will start to happen.
