MP3 - MPEG-1 Audio Layer III

MP3 is the compressed audio format that made digital music portable. Its full name is MPEG-1 Audio Layer III. It grew out of years of research at the Fraunhofer Institute for Integrated Circuits in Germany, with university and industry partners, and was published as part of the MPEG-1 standard by the Moving Picture Experts Group in the early 1990s. Layer III was the most sophisticated of that standard's three audio layers, and it shrank a song to a size a dial-up connection could actually move. Within a few years the extension had become shorthand for a music file, and it still is.

How an MP3 stores sound

MP3 is lossy. That means the encoder permanently throws away part of the recording to make the file smaller, and no player can put it back. It works by cutting the audio into short overlapping blocks and turning each block from a waveform into a set of frequencies. A psychoacoustic model then judges which of those frequencies a human ear would miss anyway, usually because a louder sound nearby masks them, and gives those far fewer bits. What comes out is not identical to the source, only close enough that most people hear no difference.

The file itself is a simple chain of frames. Each frame holds a fixed number of samples, 1152 in the common MPEG-1 flavour, which is about 26 milliseconds of sound at CD sample rates. Every frame begins with its own small header giving the bitrate, sample rate and channel mode. That is why you can cut an MP3 anywhere and the rest still plays, and why variable bitrate files work at all: the encoder can spend more data on a dense passage and almost none on silence.

What metadata an MP3 carries

Audio frames have no room for a song title, so MP3 metadata lives in a separate tag attached to the file. There are two kinds, and one file can have both. ID3v1 is the older one: the last 128 bytes of the file, with fixed-width slots for title, artist, album, comment, a four-digit year and a single genre number. Anything longer is simply cut off. ID3v2 sits at the front and is far richer. It is built from frames named by four letters, which a metadata report usually translates into words for you.

The fields you see most often are TIT2 (title), TPE1 (artist), TALB (album), TRCK (track number), TCON (genre), a recording date, COMM (free-text comment), APIC (the embedded cover image) and TXXX, which lets any program invent a named field of its own. Ripping and tagging software fills most of these in, so what you find often describes the software as much as the music.

Some of it is personal. TENC records who or what encoded the file and TSSE the encoder settings, so a report can name the exact program used. Comment frames sometimes hold a real name, an email address or a note the sender forgot about. TOWN and TPUB name a licensee and a publisher, PRIV frames hold private data written by one particular application, and a track bought from a store can carry an account or order identifier. Most encoders also write a hidden Xing or LAME header inside the first frame, revealing the encoder version, its settings and the measured peak level. MP3 has no standard field for a location or a device, so a recording made on a phone usually gives away the app and a timestamp rather than a place.

When to use MP3

Choose MP3 when you want audio that plays everywhere with no questions asked, from a car stereo to a decade-old phone. Nothing else has that reach. If size matters more than compatibility, AAC and Opus reach similar quality in a smaller file, because both were designed later with better tools. If the audio matters more than the size, FLAC and WAV keep every sample intact. And remember that lossy compression compounds: editing an MP3 and saving it again re-encodes it, so work from the best original you have and export to MP3 last.