Encodec
| Audio Compression Technique | |
|---|---|
| Developer | |
| Release Date | October 25, 2022 |
| Type | Audio compression |
| Bitrate Range | 1.5 kbps to 12 kbps |
| Algorithm | Neural network-based (Encodec) |
Encodec is an audio compression technique developed by Facebook. The technique uses artificial intelligence to achieve highly efficient compression rates, approximately 10 times better than traditional MP3 at 64 kbps, while maintaining audio quality. This method is designed to work with 48 kHz sampled stereo audio (CD quality), which is the standard for music distribution.
Overview[edit]
The Encodec technique involves training a neural network end-to-end to compress and decompress audio files. The system comprises three parts:
Encoder: Transforms uncompressed data into a higher dimensional and lower frame rate representation. Quantizer: Compresses the representation to the targeted size, while retaining the most important information to rebuild the original signal. Decoder: Converts the compressed signal back into a waveform as similar to the original as possible, using discriminators to improve perceptual quality. The technique achieves state-of-the-art results in low bit rate speech audio compression (1.5 kbps to 12 kbps), as evaluated by human annotators. The model encodes and decodes audio in real time on a single CPU core.
Advantages[edit]
The Encodec technique offers the following advantages:
Improved audio compression rates compared to traditional codecs like MP3, Opus, and EVS. Real-time audio encoding and decoding. Potential for further improvements in file size reduction and optimization of the trade-off between computing power and compressed audio size.
External links[edit]
This article "Encodec" is from Wikipedia. The list of its authors can be seen in its historical and/or the page Edithistory:Encodec. Articles copied from Draft Namespace on Wikipedia could be seen on the Draft Namespace of Wikipedia and not main one.
