MP3 Audio
MP3 is MPEG Audio Layer III, a lossy audio format commonly used for music and spoken recordings.
MP3's compact music files made a large collection practical on limited storage. Its name means MPEG Audio Layer III: “III” is an audio layer, not MPEG-3 video.
The format combines time/frequency analysis, perceptual precision allocation, and efficient symbol coding. Understanding a few of its tools makes the labels in a media report useful.
Audibility estimates and quantization
The encoder analyzes sections of the waveform, estimates where coding error is likely to be audible, and quantizes the signal representation. The decoder reconstructs sound from those coded values. The reconstructed PCM is an approximation of the input.
More data can permit finer precision or retain difficult details, but the choice of encoder tools and material also matters. Lossy compression follows the general process through a numerical example.
Joint stereo shares information rather than removing a side
Left and right can contain substantial common content. A mid/side model illustrates a reversible way to describe that relationship:
M = (L + R)/2; S = (L − R)/2.
L = M + S; R = M − S.
| Example input | Mid | Side |
|---|---|---|
| L=+0.8, R=+0.6 | +0.7 | +0.1 |
| L=+0.4, R=+0.4 | +0.4 | 0 |
| L=+0.4, R=−0.4 | 0 | +0.4 |
MP3's tools have their specified normalization and coding behavior; this algebra shows why shared stereo information can be described efficiently without turning everything into mono.
Joint-stereo modes can choose suitable shared representations, and MP3 also has intensity-stereo tools that make a different perceptual tradeoff. The label is consequently a family of coding choices, not proof that a channel disappeared.
Frame bitrates and the bit reservoir
CBR selects a nominal frame bitrate. VBR allows the encoder's frame-bitrate choices to vary with the material. MP3 also has a bit reservoir: data associated with one frame can use space carried earlier. Even a CBR stream therefore has more flexibility than “every passage receives exactly the same coding treatment.”
At an average 192 kbit/s, four minutes contain about 5.76 MB of encoded data before tags and overhead. At an average 320 kbit/s, the same duration uses 9.6 MB. These are size calculations, not thresholds of guaranteed transparency.
Audio frames and file metadata answer different questions
The MPEG audio frames carry sound. ID3 metadata can carry titles, artists, and artwork. Optional navigation or trimming information assists duration, seeking, and presentation boundaries. The MP3 packaging article develops those structures.
A file can therefore have correct audio but incorrect artwork, or valid tags while suffering an audio/frame problem. A frame boundary is also not automatically an independent clean decoding start when reservoir dependencies exist.
When an Emby music path converts MP3 to another lossy codec, the new encoder starts from the MP3 reconstruction. Keeping the master or the original delivered MP3 avoids making future versions from a chain of converted copies.