Audio & music glossary

Quick definitions of the technical and musical words used in Full Audio Tools's tools and tutorials, written for singers, musicians, DJs and creators rather than engineers.

Acapella: A voice heard on its own, without instruments. DJs and producers use the isolated vocal of a song (the "acapella") to build remixes and mash-ups.
AMR: Adaptive Multi-Rate, a speech codec designed for mobile networks and used by some older phone recorders. Most web browsers can't decode it.
Bitrate: How much data a file spends on each second of sound, written in kbps. More data usually means better quality and bigger files.
Bit depth: The precision of each audio sample. Higher bit depths lower the noise floor; 24-bit is common for recording and 16-bit for finished music.
BPM (beats per minute): The speed of a piece of music, counted as beats per minute. Most pop sits roughly between 90 and 130 BPM.
Camelot wheel: A DJ tool that labels the 24 musical keys 1A–12A (minor) and 1B–12B (major) so compatible keys sit next to each other.
Cents: A fine unit of pitch: 100 cents make one semitone. Small cent adjustments fix recordings that are slightly sharp or flat.
Clipping: Harsh distortion caused when a signal is pushed past the maximum digital level and the peaks of the waveform are chopped off.
Codec: The method used to encode and decode audio data. MP3, AAC, Opus and FLAC are all codecs.
Decibel (dB): A logarithmic unit for comparing levels. +6 dB doubles a signal's amplitude, and digital levels are measured in dBFS, where 0 is the maximum.
Formant: A natural resonance of the vocal tract that gives a voice its recognisable character. Formant-preserving pitch shifting keeps voices sounding human after a key change.
Gain staging: Setting the level at each step of recording and processing so the signal stays well above the noise without ever clipping.
Harmonic mixing: Choosing or transposing tracks so that their keys are compatible, making DJ transitions sound smooth instead of clashing.
Instrumental: A version of a song without vocals, also called a backing track or karaoke track.
Key: The home note and scale a song is built around, such as G major or E minor. There are 12 major and 12 minor keys.
Limiter: A level control that prevents audio from going above a ceiling by briefly turning down only the loudest moments.
LUFS: A standard measurement of how loud audio sounds to people over time, used by streaming services and broadcasters to keep volume consistent.
M4A: An audio file in the MP4 container, usually holding AAC audio. It's the format of iPhone voice memos and many music downloads.
Mono: Audio with a single channel. Ideal for speech because it halves the data compared with stereo.
Neural network: A machine-learning model made of many connected layers that learns patterns from examples. In audio, neural networks separate stems and recognise instruments.
Normalize: Raising or lowering a whole file by one fixed amount so its loudest point (or its average loudness) hits a target level.
Octave: The distance between one note and the next note with the same name, twelve semitones apart. Going up an octave doubles the frequency.
ONNX: Open Neural Network Exchange, a portable file format for AI models that lets them run in different environments, including web browsers.
Opus: A modern, open audio codec that sounds good at very low bitrates. It's used by many voice-message and calling apps, often in .opus or .ogg files.
Peak level: The highest instantaneous level in a recording, measured in dBFS. It shows how much headroom remains before clipping.
Pitch: How high or low a sound is perceived, set mainly by its fundamental frequency.
Relative key: A major and a minor key that share exactly the same notes, such as C major and A minor. They mix together naturally.
Room tone: A few seconds of "silence" recorded in a space, capturing its natural background sound. Editors use it to fill gaps seamlessly.
Sample rate: How many times per second sound is measured when it's digitised. 44.1 kHz and 48 kHz are standard for music and video.
SDR (signal-to-distortion ratio): The standard score for stem-separation quality, in decibels. Higher numbers mean less bleed and fewer artefacts.
Semitone: The smallest step between notes in Western music, from one piano key to the very next. Twelve semitones make an octave.
Stem: One part of a song (vocals, drums, bass, or everything else) as its own audio file. All stems together make the full song.
Tempo: The speed of music, usually given in BPM. Time-stretching changes tempo without affecting pitch.
Tessitura: The part of a singer's range where their voice sounds best and feels most comfortable for long periods.
Time-stretching: Making audio longer or shorter, and so slower or faster, while keeping its pitch the same.
Transient: The short, sharp start of a sound, such as a drum hit or a plucked string. Keeping transients intact keeps audio punchy.
Transpose: To move a whole piece of music up or down by the same number of semitones, changing its key but not its melody.
Vocal range: The span from the lowest to the highest note a person can sing, often grouped into voice types such as soprano, alto, tenor and bass.
Waveform: The graph of an audio signal's level over time, as shown in audio editors. Loud passages appear taller.
WebGPU: A web standard that lets sites use your graphics card for heavy computation, such as running AI separation models without uploading anything.