Meta has introduced the Meta Low Bitrate, also known as MLow, with the aim of improving the quality of low bitrate audio data in real-time voice communication between devices, especially in slow connection environments.
The challenge of real-time communication services is to maintain the highest possible data quality. While general voice calls typically have a bitrate of 768 kbps, the latest converters can compress data down to levels of 25-30 kbps. However, this reduction in size comes at the expense of quality. Thus, the trade-off involves bitrate, quality, and complexity of the converter.
Meta acknowledges that the development of new converters is not frequent. The primary tool widely used by the platform is Opus, an open-source converter introduced in 2012, which serves as the foundation for developing new converters.
Machine Learning (ML) has significantly supported the enhancement of new audio converters. Meta previously unveiled Encodec in 2022, which improves the quality of low bitrate audio using ML. Nonetheless, Encodec requires high processing resources and performs best on high-spec smartphones.
Developed in late 2021, MLow has been found, on average, to offer better audio quality than Opus 2 while utilizing less than 10% of processing resources. MLow operates by separating audio encoding into high and low-frequency segments, then re-encoding them before transmitting the data, requiring the recipient to decode the audio into two segments as well.
Currently, MLow has been implemented for voice calls on Messenger and Instagram, and it has recently started being used on WhatsApp for a large number of voice calls.
Source: Meta
TLDR: Meta introduced MLow to enhance audio quality in low bitrate communication, utilizing ML technology and achieving better quality with less processing resources than traditional converters.
Leave a Comment