Audio glossary

Short, plain-English definitions of the audio terms used across All Audio Tools. Jump to a letter, or use your browser's search (Ctrl/⌘ + F).

AAC (Advanced Audio Coding): A lossy audio codec that succeeded MP3, giving similar quality at lower bitrates. It's used for iPhone recordings, Apple Music and most MP4 video soundtracks, usually in an .m4a file.
Acapella: A vocal performance without instrumental accompaniment. In production, it usually means the isolated vocal track of a song, used for remixes and practice.
AIFF: Audio Interchange File Format: Apple's uncompressed audio format, equivalent in quality to WAV.
Bit depth: The number of bits used to store each audio sample. It sets the dynamic range: about 96 dB for 16-bit and about 144 dB in theory for 24-bit.
Bitrate: The amount of data used per second of audio, in kilobits per second (kbps). Higher bitrates generally mean better quality and larger files. File size in MB ≈ kbps × seconds ÷ 8,000.
Bleed: Unwanted traces of one sound in a track meant for another, such as drums audible in a separated vocal stem or a second instrument picked up by a microphone.
CBR (constant bitrate): Encoding that uses the same bitrate throughout a file, making sizes predictable. Compare VBR.
Channel: An independent stream of audio. Mono has one channel, stereo has two (left and right), and surround formats have more.
Clipping: Distortion that occurs when a signal tries to exceed the maximum level (0 dBFS in digital audio), so the tops of the waveform are cut flat. It sounds harsh and crackly.
Codec: Short for coder–decoder: the method used to compress and decompress audio, such as MP3, AAC, Opus or FLAC.
Container: The file format that holds encoded audio (and sometimes video and metadata), such as MP4/M4A, OGG or WebM. The same codec can live in different containers.
Crossfade: A transition in which one sound fades out while the next fades in, overlapping. Equal-power crossfades keep the overall loudness steady.
dBFS (decibels relative to full scale): The unit for digital audio levels. 0 dBFS is the highest possible level; all other levels are negative numbers below it.
Dynamic range: The difference between the quietest and loudest parts of a recording, or the range a format can represent, measured in decibels.
Fade in / fade out: A gradual increase in level at the start or decrease at the end of a clip, used to avoid abrupt starts, stops and clicks.
FLAC: Free Lossless Audio Codec: compresses audio to roughly half to two-thirds of its WAV size with no loss of quality.
Formant: A resonant frequency band of the human voice that gives each voice its character. Preserving formants during pitch shifting keeps singers from sounding like chipmunks or giants.
Gain: An increase or decrease in signal level, usually in decibels. +6 dB roughly doubles the signal's amplitude.
Headroom: The space between a signal's peak level and the maximum (0 dBFS). Headroom protects against clipping during recording and processing.
Hertz (Hz) and kilohertz (kHz): Cycles per second, the unit of frequency. Human hearing spans roughly 20 Hz to 20 kHz.
Key (musical): The tonal centre and scale a piece of music is built on, such as A minor or E♭ major. Transposing moves a song to a different key.
Limiter: A processor that stops a signal from exceeding a set ceiling by briefly reducing gain on peaks. A look-ahead limiter sees peaks in advance so it can react smoothly.
Lossless: Compression or storage that preserves every bit of the original audio, as in WAV, FLAC and ALAC.
Lossy: Compression that permanently discards detail judged inaudible, to make files much smaller, as in MP3, AAC, Vorbis and Opus.
LUFS: Loudness Units relative to Full Scale: a measure of perceived loudness over time (ITU-R BS.1770). It's used by broadcasters and streaming platforms to even out volume between programmes.
Mono: Single-channel audio. Speech is often published in mono because it halves the data needed without an audible loss.
MP3: MPEG-1 Audio Layer III, the most widely compatible lossy audio format, supporting bitrates up to 320 kbps.
Normalization: Adjusting a file's overall level to a target. Peak normalization targets the loudest sample; loudness normalization targets an average loudness in LUFS.
Opus: An open, highly efficient lossy codec, standardised in 2012, that excels at low bitrates. It's widely used for voice calls and voice messages.
PCM (pulse-code modulation): The standard way of representing uncompressed digital audio as a sequence of samples, used in WAV, AIFF and CDs.
Pitch shifting: Changing the pitch of audio, measured in semitones and cents, either with or without changing its duration.
RMS level: Root-mean-square level: an average of a signal's energy over time, closer to perceived loudness than the peak level.
Sample: (1) A single measurement of a sound wave in digital audio. (2) A short recorded sound reused in a new piece of music.
Sample rate: How many samples are taken per second, in Hz. It can represent frequencies up to half its value, so 44.1 kHz covers up to about 22 kHz.
Semitone: The smallest interval in Western music, one piano key to the next. Twelve semitones make an octave, and one semitone is 100 cents.
Source separation: The process of splitting a mixed recording into its component sources, such as vocals, drums and bass, today usually with neural networks such as HTDemucs.
Stem: A group of related tracks mixed into one file, such as all drums or all vocals. Stems played together recreate the full mix.
Stereo: Two-channel audio (left and right) that creates a sense of width and placement.
Time-stretching: Changing the duration or tempo of audio without changing its pitch, using algorithms such as phase vocoders or WSOLA.
True peak: The highest level the analogue waveform reaches between digital samples. It can exceed the highest sample value, so delivery specs often require true peaks of −1 dB or lower.
VBR (variable bitrate): Encoding that uses more bits for complex passages and fewer for simple ones, often improving quality per megabyte. Compare CBR.
WAV: Waveform Audio File Format: a widely supported uncompressed format storing PCM audio, standard in recording and editing.
WebGPU: A modern browser technology that lets web pages run computations on the device's graphics card. It's used here to run AI audio models locally.