Google launches Lyra, its open-source audio codec that relieves bandwidth

RAW audio compressed at just 3 kb per second: that’s the pretty incredible promise of Lyra, a new audio codec that Google has made open-source today. In a world that has flocked to phone calls and conferences, the “bill” for bandwidth taken up by voice transit is high. And participates in network congestion.

Lyra is a codec similar in appearance to the others in its initial approach – slicing up small 40ms chunks compressed by the sender and uncompressed by the receiver – but differing in method. Instead of focusing on pure signal processing, Lyra relies on a decoder that uses a particular model of machine learning: the generative model. Basically, the decoder is more “intelligent” because it is trained to reconstruct a quality audio signal. Its strength in reducing its bandwidth is that it relies on the receiving terminal to reconstruct part of the signal and not on an excess amount of data to be transmitted.

According to the Google blog, the audio quality would be equivalent to other codecs but would take less bandwidth. Rather than seeing its codec as a panacea, Google presents it as an ideal tool for freeing up bandwidth during degraded conditions (large-scale disasters such as fires), for networks with strong user growth without adding new ones. infrastructure or for archiving a large amount of audio voice content.

Also to discover in video:

Also to discover in video:

Developed in C ++, Lyra and its APIs are available on the GitHub repository that Google dedicates to the project.

Source : VentureBeat


Please enter your comment!
Please enter your name here