Meta announces 'MLow', a low-bitrate, high-quality audio codec developed for smartphone calling apps
Meta, the developer of
MLow: Meta's low bitrate audio codec - Engineering at Meta
https://engineering.fb.com/2024/06/13/web/mlow-metas-low-bitrate-audio-codec/
When making voice calls using an RTC app on a smartphone, the codec is important. The bit rate of the audio captured in a typical call is approximately 768 kbps for mono, 48 kHz sampling rate, and 16-bit bit depth. However, in order to make a comfortable call on a smartphone, it is necessary to encode and compress the audio to reduce the bit rate.
At the time of writing, the codec most commonly used in RTC apps is Opus. Opus is an open source codec released in 2012 that is free and achieves low latency of 5 milliseconds.
'Opus', a lossy audio compression codec that achieves ultra-low latency of 5 milliseconds for free, what's so great about it?
According to Meta, which develops several RTC apps that use the Opus codec, in actual usage, network connections can be poor in some parts of the communication. When network quality deteriorates, the app automatically lowers the codec bit rate to avoid network congestion , but this inevitably leads to a deterioration in sound quality. In particular, video calls have become common in recent years, so when network quality deteriorates, the bandwidth for audio data becomes narrower, which tends to further reduce the bit rate.
Below is an example of audio captured from a typical call.
And the network environment worsens, and the audio is encoded at a low bit rate of 6 kbps with Opus. Compared to the original audio, you can see that the quality is far from clear and natural.
Therefore, Meta developed MLow as a new voice codec for RTC. MLow has the characteristic that it has twice the voice quality than Opus, but has 10% lower computational cost than Opus.
MLow splits the input audio into low and high frequency bands, encodes each separately, and each band uses shared information to improve compression efficiency, then the range encoder compresses it further to produce the payload, and the decoder reverses the process to produce the output audio signal.
The graph below compares the score (vertical axis) of the voice quality test '
When comparing the original audio, Opus, and MLow, it can be seen that the sound quality of Opus is significantly degraded, while MLow maintains a sound quality that is quite close to the original audio.
| Original audio | Opus 6kbps | MLow 6kbps |
According to Meta, MLow has already been fully implemented in Instagram and Messenger calling features, and is gradually being rolled out to WhatsApp. Meta commented, 'We're committed to making high-quality voice calling easier for all our users, and look forward to sharing more information in the future.'
◆ Forum is currently open
A forum related to this article has been set up on the official GIGAZINE Discord server . Anyone can post freely, so please feel free to comment! If you do not have a Discord account, please refer to the account creation procedure article to create an account!
• Discord | 'What device or app do you use for voice calls? Are you satisfied with the current call quality?' | GIGAZINE
https://discord.com/channels/1037961069903216680/1251113199604858901
Related Posts:
