Opus, New codec: goodbye MP3?
There is a lot of talk on the web that the new Opus audio codec can replace the MP3 format. Read about how to improve the new compression algorithm in our article
Individual man-made discoveries turn out to be inventions so practical that they remain in everyday life for a long time. For example, the MP3 digital audio compression standard is about 20 years old, which is more than a long time by computer technology standards. During these twenty years, there have been many discoveries and technological advances. But for digital audio, oddly enough, little has changed so far. MP3 has found its way into all kinds of devices: smartphones, portable players, DVD players, watches, and other electronic devices.
Why is that? After all, lossy encoding is an inevitable degradation of sound quality. To the layman it may seem that MP3 has had alternatives for a long time: FLAC, APE and other algorithms for compressing audio data with the possibility of an identical restoration of the waveform after decoding. The judgment that the advent of lossless audio compression algorithms will compete with MP3 in all respects is very superficial. In addition to the sound quality that fans of music collections are so concerned with, there are many other objective reasons why MP3 cannot be forgotten and replaced by the principles of lossless compression.
First of all, because lossy audio encoding formats are used not only for music but also for voice over the Internet. The trump card of MP3 and other lossy compression mechanisms is the efficient use of transmission channels. To organize IP telephony, it is necessary to provide intelligible voice to as many subscribers as possible. In this case, the sound quality takes a back seat. Furthermore, the possibility of “instantaneous” decoding of the stream is very important, without which the synchronized exchange of information is difficult. In this case, the use (even theoretically) of lossless compression algorithms would lead to severe delays and interactive communication would simply be impossible.
However, MP3 is not without its drawbacks. It’s no secret that a low bit rate “devours” the details of the sound, endowing it with a whole host of unpleasant artifacts: overtones, hiss and timbres, all kinds of distortions. When using MP3 in IP telephony, there are long delays due to the need for additional data buffering.
⇡ # Opus: a new word in digital sound
The new open Opus codec lacks the most serious drawbacks of MP3, while retaining all the advantages of the “popular” codec and even multiplying them.
The Opus structure enables it to effectively deal with sound artifacts. For this, a multi-stage audio signal processing architecture has been proposed. The main argument that speaks in favor of the use of the new codec for IP telephony is the low delay.
The main work in creating a unique compression algorithm was done by several people: Jean-Marc Valin (Xiph.Org, Octasic, Mozilla Corporation), Koen Vos (Skype) and Timothy B. Terriberry (Xiph.Org, Mozilla Corporation) ). Not without the ubiquitous Google: According to the creators of Opus themselves, the internet giant provided significant support in the development and testing of the codec.
One of the creators of the Opus codec – Jean-Marc Valin
The new codec engine is based on two independent standards proposed by the Xiph.Org Foundation and Skype Technologies SA (owned by Microsoft). The new codec is a hybrid solution that combines the technologies of the CELT (Constrained Energy Lapped Transform) and SILK codecs. The latter is used to implement communication in Skype.
⇡ # How Opus works
The working principle of the codec is not new, but original and, most importantly, it allows to obtain a very good result in the output. The received signal is selectively encoded SEDA or CELT.
The first engine (SILK) is used for voice compression, as well as in cases where it is required to efficiently use the bandwidth of the communication channel. The processed audio signal is analyzed by the codec to detect the presence of human speech. The speech components are separated from other sounds, after which the codec analyzes the frequency response of the sound, reducing the resolution of data containing speech information, i.e. speech. Then Opus examines the noise present and optimizes the signal for a specific bit rate. The codec then converts the signal using a pre-filter. Using speech frames, the prediction module
Well, this article sure opened my eyes to the world of psychoacoustics and Opus encoding. Never thought about the impact of bitrates on audio quality! Can’t wait to try out the suggested optimizations. Kudos!
Hey, fantastic read! However, I was expecting a bit more depth on the real-life applications. Could you perhaps share more examples or case studies? Cheers!
This article is a gem! Finally, someone addressing the future trends in audio encoding. Super insightful and thought-provoking. Can’t wait to see where the industry is headed.
Great stuff! Tried the suggested Opus encoding steps, and the audio quality is noticeably better. Huge thanks for the practical tips. Keep ’em coming!
While I appreciate the overview, I wish there was more detail on the psychoacoustic research. Maybe a dedicated section in the future? Overall, solid content.