Meta Introduces EnCodec Technology 10 Times More Compression Than MP3 With The Same Quality

Since the introduction of the MP3 codec around 30 years ago, it feels like little has changed in the field of audio compression. But now the development could come to a point where a significantly better technology could offer an alternative.

At least that’s what the new “EnCodec” technology that the social media group Meta has presented suggests. According to the company, this is able to compress audio data up to ten times smaller with the same quality. The company’s developers primarily want to use it to improve the quality of voice transmissions, for example via messenger platforms. But they also emphasize that it can also be used with music without any problems.

In order to achieve decisive improvement, a three-stage process using AI technologies is used. In the first step, the uncompressed data is translated into a lower bit rate. The result is then scaled down to the desired target size by an AI called a “quantizer”.

More computing power

The signal compressed in this way is then stored or transmitted and has a very small data volume. If it is then to be played back by the recipient, another AI neural network ensures that the audio information is restored. Here, a normal CPU is sufficient to achieve real-time conversions, it said.

This means that the necessary use of computing power increases, especially during playback. At the time when the MP3 format was invented, this would have been unacceptable because the resources available on computers and even more so on other playback devices were extremely limited. Today, on the other hand, the built-in chips are so powerful that the additional effort should hardly be significant.

The technology is still in the research phase and the algorithms are therefore in an early form. So the public can’t yet be convinced of exactly what EnCodec can do. That should change soon given that Meta has now presented the development.