An audio codec is a technology for encoding and decoding digital audio, with the term “codec” derived from “coder-decoder” or “compressor-decompressor”. An encoder converts audio into a specified digital representation, often compressing it to reduce storage or transmission requirements, while a decoder reconstructs audio samples or a playable representation from the encoded data.
Some codecs are lossless, meaning the original digital audio samples can be reconstructed exactly, while others are lossy, meaning some information is deliberately discarded to achieve greater compression. Examples include MP3, AAC, Opus, Vorbis, FLAC and ALAC. A codec is not necessarily the same thing as a file format or container: a container can hold audio encoded with a particular codec along with metadata and other streams.
Audio codecs exist because uncompressed audio can require substantial amounts of data. For example, stereo PCM at 44.1 kHz and 16 bits per sample requires 1,411,200 bits per second before considering file-container overhead. A codec can reduce this requirement dramatically by exploiting statistical redundancy, perceptual characteristics of hearing, or both. Lossless codecs find compact mathematical representations that preserve every sample, while lossy perceptual codecs identify information that can be removed or represented less precisely because it is less likely to be perceived under relevant listening conditions. Compression therefore reduces storage and network requirements but introduces different trade-offs depending on the codec and settings.
Lossy codecs commonly use psychoacoustic principles. Human hearing does not perceive every component of a complex sound equally, and strong sounds can affect the audibility of weaker sounds occurring at nearby frequencies or times. A perceptual encoder can analyse the audio, estimate which components are more or less important to perception, and allocate bits accordingly. MP3 and AAC use sophisticated forms of transform coding and psychoacoustic modelling, although their internal designs differ. The encoder may discard information judged to have low perceptual importance and represent important components more accurately. At lower bitrates, the compromises become more substantial and can produce artefacts such as pre-echo, smearing, high-frequency degradation or other audible distortions.
Lossless codecs use a fundamentally different strategy. They do not intentionally remove audio information. Instead, they exploit predictable patterns and statistical redundancy so that the encoded representation occupies fewer bits than the original PCM data while retaining enough information to reconstruct every sample exactly. FLAC is a prominent example. The encoder can predict samples from neighbouring samples and encode the residual error efficiently, while additional techniques represent repeated or statistically predictable information compactly. When decoded, the original sample sequence is recovered bit-for-bit. Lossless compression generally achieves less size reduction than lossy perceptual compression, but it is valuable when preservation of the exact digital source is important, such as archival storage, production workflows and high-fidelity distribution.
Codec choice affects compatibility, computational requirements, quality, latency and file size. Some codecs are widely supported by operating systems, browsers and hardware, while others may require specialised software. Encoding and decoding complexity matters for battery-powered devices, embedded systems and real-time communications. Low-latency applications may favour codecs designed for rapid processing and packet-loss resilience, whereas music archiving may prioritise lossless preservation. Streaming services may select codecs according to device support, licensing considerations, network efficiency and desired quality. Consequently, there is no universally optimal codec for every application; the appropriate choice depends on the characteristics of the content and the constraints of the delivery system.
Containers are frequently confused with codecs. MP4, Matroska and Ogg are examples of container formats capable of holding encoded media streams, while AAC, H.264 and FLAC identify particular encoding technologies. A file extension may indicate a container rather than the exact codec inside it. Metadata can also be stored alongside the encoded audio, including track title, artist, album information, artwork and technical parameters. Separating container from codec is essential when diagnosing compatibility problems: a player may support a container but not a particular codec inside it, or it may support the codec but not the container. Media software therefore usually identifies both the outer format and the underlying audio encoding.
An audio codec can operate in offline file compression or real-time communication. In a stored music file, an encoder can spend considerable processing time analysing the entire track and optimising compression. In a telephone or video-conferencing application, encoding and decoding must happen quickly, and the codec must often tolerate packet loss and variable network conditions. Real-time codecs can use small frames to reduce latency and may include mechanisms such as packet-loss concealment. The trade-off between bitrate, delay, complexity, robustness and quality becomes particularly important in interactive communication. Thus, “audio codec” encompasses technologies with very different design priorities even though all perform the broad task of representing audio for storage or transmission.
The quality of a codec cannot be judged solely by its name. Encoder implementation, bitrate, sample rate, channel configuration, source material and decoding environment all influence results. Two files using the same codec can sound different because they were encoded with different settings or source quality. Transcoding from one lossy codec to another can introduce additional degradation because information already discarded by the first encoder cannot be restored by the second. Conversely, decoding a lossless file and re-encoding it losslessly preserves the digital samples exactly. Understanding codecs therefore requires distinguishing the codec algorithm from the encoding parameters and the quality of the original source.
