Meta has developed an AI-powered audio compression algorithm called EnCodec that claims to achieve 10x better compression rates than traditional MP3 format while maintaining high audio quality.

EnCodec uses a three-part system to compress audio:

• An encoder that converts uncompressed audio into a lower frame rate representation
• A quantizer that compresses the signal while preserving essential information
• A decoder that reconstructs the audio in real-time using neural networks

The system employs discriminators in a "cat-and-mouse game" where the compression model tries to generate samples that fool the discriminators, resulting in better perceptual quality. Meta researchers claim this is the first neural network application for compressing 48 kHz stereo audio - slightly higher than CD quality at 44.1 kHz.

Primary applications include:

• Improved voice calls over poor network conditions
• Enhanced metaverse audio experiences without increased bandwidth requirements
• High-quality audio delivery across varying network conditions

While still in research phase, EnCodec represents a significant advancement in audio compression technology that could transform how we transmit and consume audio content.

Related Articles

Previous Articles