Reference

Audio AI Glossary

Plain-language definitions of the terms we use across stem separation, speech AI, music rights, and broadcast delivery. Search for a term or browse A–Z.

All #A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

1–20 of 188

A six-channel surround layout: left, center, right, two surrounds, and LFE.

An immersive bed layout adding four height channels to 7.1.

7.1.4 Immersive

An eight-channel surround layout adding rear channels to 5.1.

A project-interchange format for moving edits and audio between production tools.

Advanced Authoring Format

A singing vocal recording presented on its own, with no instrumental backing.

Vocal Track; A Cappella; Acappella

ADR (Automated Dialogue Replacement)

The film/TV process of re-recording dialogue in a studio to replace or improve location audio.

AEC (Acoustic Echo Cancellation)

Removing the echo of a device's own speaker output from its microphone input during a call.

Echo Cancellation

A standard that lets professional audio-over-IP systems from different vendors interoperate on one network.

Automatic Gain Control, circuitry that keeps audio at a consistent level.

Automatic Gain Control

An uncompressed, lossless audio file format developed by Apple, comparable to WAV.

Unwanted sounds or distortions introduced into audio by processing, compression, or separation.

ASR (Automatic Speech Recognition)

Technology that automatically converts spoken audio into written text.

Speech-to-Text; Transcription

The North American loudness recommendation implementing the CALM Act.

Removing unwanted noise, hum, echo, and artifacts to improve a recording's quality.

Audio Restoration

Audio Description

Narration added to video that describes key visual elements for blind and low-vision audiences.

Audio Fingerprinting

A technique that creates a unique digital signature of audio so it can be identified or matched.

Content Recognition; ACR

Audio Forensics

The analysis, enhancement, and authentication of audio recordings, often for legal or investigative use.

Audio Redaction

Removing or muting sensitive information, such as personal data, from an audio recording.

Audio Source Separation

Splitting a mixed recording into its individual components, such as vocals, instruments, dialogue, or effects.

Source Separation; Stem Separation; De-mixing; Demixing

Automatically labeling audio with descriptive metadata such as instruments, genre, mood, or sound events.

Metadata Enrichment

Audio Watermarking

Embedding an inaudible identifier into audio for tracking or measurement.

When a user interrupts a voice assistant mid-response and the system stops to listen.

The number of bits used to represent each audio sample; higher bit depth means greater dynamic range.

The amount of data used to represent audio per second (in kbps), affecting file size and quality.

Unwanted sound from one source captured by a microphone meant for another, blending sources together.

Burned-in Captions

Captions or subtitles rendered permanently into the video frame.

Hardcoded Captions; Open Captions

Broadcast Wave, a WAV variant with metadata for professional workflows.

Broadcast Wave Format

US law requiring TV commercials to match the loudness of surrounding programming.

Commercial Advertisement Loudness Mitigation Act

Displaying on-screen text of spoken dialogue and relevant sounds, in the same language as the audio.

Closed Captioning

Catalog Preparation

Processing a large audio library in bulk, e.g. separating, cleaning, and tagging it, to ready it for a new use.

Data Preparation

The legacy analog-era closed caption standard, also called line-21 captions.

EIA-608; Line 21 Captions

The digital television closed caption standard for ATSC broadcasts.

Channel Mapping

The assignment of audio content to specific channels in a delivery file.

Track Layout; Audio Configuration

Distortion caused when an audio signal exceeds the maximum level a system can handle.

Closed Captions

Viewer-toggleable text of dialogue and audio, transmitted separately from the picture.

Cocktail Party Problem

The challenge of isolating one sound source, like a single voice, from a mix of many competing sounds.

An algorithm that encodes and decodes audio or video, usually with compression.

Confidence Scores

A value indicating how certain an AI model is about a given prediction, such as a transcribed word or a detected label.

Confidence Score

Conversational AI

AI systems that understand spoken or written language and respond in a natural, back-and-forth way.

Voice AI; Voice Agent

Crowd Sweetening

Augmenting live crowd audio with pre-recorded or synthesized atmosphere to enhance a broadcast.

A document logging every piece of music used in a production, with timing and rights details.

Music Cue Sheet

A widely used audio-over-IP networking technology for professional audio.

Dante Audio-over-IP

Decibels relative to digital full scale, the reference for digital audio levels.

Decibels Full Scale

Decibels True Peak, the unit for true-peak level measurement.

Decibels True Peak

Removing unwanted background noise from a recording while preserving the desired sound.

Denoising; Background Noise Removal

Reducing echo and room reverberation from a recording to make it sound clearer and closer.

Dereverb; Reverb Removal

Metadata in Dolby streams indicating average dialogue loudness for playback leveling.

Dialogue Normalization

Dialogue Isolation

Separating spoken dialogue from music, effects, and noise in a mixed soundtrack.

Dialogue Separation; Speech Isolation

Discrete Channels

Independent audio channels kept separate rather than matrix-encoded.

Any unwanted alteration of an audio signal that degrades its quality, such as clipping or harshness.

DME (Dialogue, Music & Effects)

The three core audio layers of film/TV content (dialogue, music, effects) kept as separate stems.

A next-generation audio codec for ATSC 3.0 and streaming, supporting immersive and personalized audio.

An object-based immersive audio format adding height and precise placement.

Dolby Digital (AC-3)

The multichannel codec used for broadcast and disc surround sound.

Dolby Digital Plus (E-AC-3)

An enhanced, more efficient successor to Dolby Digital for streaming and broadcast.

E-AC-3; DD+; EAC3

A broadcast codec for carrying multichannel audio over stereo-width paths.

Combining a multichannel mix down to fewer channels, such as 5.1 to stereo.

Fold-down; Downmixing

An audio signal with no added effects or reverb; the raw, unprocessed sound.

DSP (Digital Signal Processing)

Manipulating audio through digital mathematical operations such as filtering, EQ, or compression.

A family of multichannel audio codecs used in cinema, disc, and streaming.

Replacing the original spoken dialogue in content with a translated or re-recorded voice track.

Dynamic Ad Insertion

Server-driven insertion of targeted ads into live or on-demand streams.

Dynamic Range Compression

Reducing the level difference between loud and quiet audio.

Compression; DRC

The European loudness standard, targeting -23 LUFS integrated.

The European broadcast subtitle exchange format.

STL; EBU Tech 3264

Effects Separation

Isolating sound effects from the other elements of a mixed soundtrack.

Embedded Audio (SDI)

Digital audio multiplexed into an SDI video signal.

SDI Embedded Audio

Detecting when a speaker has finished talking so a system can respond or stop recording.

Speech or audio captured at a distance from the microphone, where noise and reverb are bigger problems.

Far-field Speech

A lossless audio format that compresses files without any loss of quality.

Forced Narrative

Subtitles shown automatically for foreign dialogue or on-screen text, even when captions are off.

Forced Subtitles; Forced Subs

Fourier Transform / STFT

The math that breaks a signal into its component frequencies; the short-time version (STFT) tracks them over time.

STFT; Short-Time Fourier Transform

A short, fixed-length segment of audio that processing and analysis work on, one piece at a time.

Frequency-Domain

A representation of audio by its frequency content rather than how it changes over time.

The smallest written unit of a language, such as a letter or letter group representing a sound.

The verified, correct reference data used to train and measure the accuracy of AI models.

A metadata standard that stores title, artist, album, and other details inside audio files like MP3.

A standardized master package format for distributing finished content in multiple versions.

Interoperable Master Format

Immersive Audio

Audio that surrounds the listener in three dimensions, including height, for a more enveloping experience.

A standardized TTML profile for subtitle and caption interchange across platforms.

Internet Media Subtitles and Captions

Running a trained AI model on new input to produce a result, as opposed to training the model.

A version of a song with the lead vocals removed, leaving only the instrumentation.

Integrated Loudness

The single loudness value for an entire program, measured over its full duration.

Program Loudness

Intelligibility

How clearly speech can be understood in a recording, especially amid noise or distortion.

Speech Intelligibility

The International Standard Recording Code, a unique global identifier for a specific sound recording.

International Standard Recording Code

The international standard algorithm for measuring loudness and true peak.

Variation in the arrival timing of audio or data packets over a network, which can cause glitches.

A small buffer that holds incoming audio packets to smooth out timing variation and prevent dropouts.

The low-frequency effects channel, the '.1' in surround formats.

Low-Frequency Effects; .1 Channel

A processor that hard-caps audio level to prevent it exceeding a ceiling.

Peak Limiter; Brickwall Limiter

The alignment of audio with the corresponding lip movements on screen.

A/V Sync; Audio-to-Video Sync

ATSC term for loudness measurement, numerically identical to LUFS.

Adapting content for a different language or region, including translation, dubbing, and captioning.

Loudness Normalization

Adjusting content to hit a consistent target loudness across programs and platforms.

Loudness Management

LRA (Loudness Range)

A measure of how much loudness varies across a program.

A matrix-encoded stereo fold-down that preserves surround information.

Left total/Right total; Lo/Ro

Standardized unit for measuring perceived loudness of audio, relative to digital full scale.

LKFS; Loudness Units Full Scale

Lyric Alignment

Synchronizing a song's lyrics to its audio with precise, word-by-word timing.

Forced Alignment

Lyric Transcription

Converting the sung words of a song into written text.

M&E (Music and Effects)

A mix containing all of a production's music and sound effects but no dialogue, used for localization.

A new track created by blending elements, such as vocals and instrumentals, from two or more songs.

The final stage of audio production, polishing a mix for consistent playback everywhere.

MIR (Music Information Retrieval)

The field of extracting information from music, such as tempo, key, chords, or instruments.

The combined result of blending multiple audio sources into a single recording.

A feed that removes a participant's own audio to prevent echo.

A trained AI system that takes input, such as audio, and produces an output, such as separated stems.

Audio with a single channel, where all sound is combined into one signal with no left/right separation.

MOS (Mean Opinion Score)

A subjective audio-quality metric based on average human listener ratings, typically on a 1-to-5 scale.

A widely used lossy audio format that shrinks file size by discarding some audio data.

A container format that can hold audio, video, and metadata together in one file.

Multi-Speaker Separation

Separating a recording with several talkers into individual tracks, one per speaker.

Speaker Separation

Music Clearance

Obtaining legal permission to use a piece of music in a specific context.

Rights Clearance

Music Detection

Automatically identifying where music is present in a piece of audio or video.

Music Identification

Recognizing the specific track playing in a piece of content.

Song Identification; Music Recognition

Removing or reducing music from a recording while keeping speech and other wanted audio.

Music Stripping

A professional container format for wrapping video, audio, and metadata.

Material Exchange Format

A protocol for sending low-latency video and audio over a local IP network.

Network Device Interface

Speech or audio captured close to the microphone, giving a strong, clean signal.

Near-field Speech

A machine-learning architecture, loosely inspired by the brain, that powers modern audio AI.

Neural Net; Deep Neural Network

Nielsen Watermark

An inaudible Nielsen code embedded in broadcast audio for ratings measurement.

Nielsen Audio Encoding; CBET

Object-Based Audio

Audio built from individual sound objects with positional data, rendered to any speaker layout.

A mobile production unit that captures and mixes audio and video on-site at a live event.

Outside Broadcast Truck; OB Van

A legacy interchange format for transferring audio and edit data between systems.

OMFI; Open Media Framework

Captions permanently rendered into the video and always visible.

Always-on Captions

Over-suppression

When noise reduction removes too much, muffling or degrading the wanted speech along with the noise.

Uncompressed digital audio, the baseline for professional and broadcast audio.

Pulse-Code Modulation; Linear PCM

PESQ (Perceptual Evaluation of Speech Quality)

An objective metric that estimates perceived speech quality, designed to approximate human listener ratings.

The timing relationship between waveforms; misalignment can cause cancellation and a thinner sound.

The smallest unit of spoken sound that distinguishes one word from another.

Portable People Meter, a device that detects watermarked audio to measure exposure.

Portable People Meter

Primary Speaker Separation

Isolating the main (foreground) speaker in a recording from other voices and background sound.

Profanity Delay

A short, intentional delay on a live broadcast that lets operators catch and remove profanity before it reaches the audience.

Broadcast Delay

Programmatic Audio

Automated, data-driven buying and insertion of audio advertising.

Programmatic Audio Advertising

The final, combined broadcast mix sent to viewers, blending commentary/dialogue, music, and effects.

Full Mix; Program Feed

The rhythm, stress, and intonation of speech, everything beyond the words themselves.

PTP (IEEE 1588)

A protocol that synchronizes clocks across networked devices with high precision, keeping audio and video in sync.

Precision Time Protocol; IEEE 1588

Processing audio fast enough to keep pace with the live signal, with imperceptible delay.

Ultra Low Latency; Live

The process of capturing sound and converting it into a stored audio signal.

Creating an improved version of an existing recording for modern playback or formats.

Creating a new version of a song by rearranging, replacing, or re-balancing its individual stems.

The persistence of sound in a space as it reflects off surfaces; too much reduces clarity.

Reverberation; Room Reverb

An open, error-resilient protocol for reliable video contribution over the internet.

Reliable Internet Stream Transport

How many audio samples are captured per second (in kHz); higher rates capture more detail.

Reusing a portion of an existing recording as an element in a new piece of music.

A secondary audio program channel, often used for alternate languages or description.

Secondary Audio Program

Scenarist Closed Caption file carrying CEA-608 caption data.

Scenarist Closed Caption

The signaling standard that marks ad breaks and splice points in a stream.

Subtitles for the deaf and hard of hearing, including non-dialogue audio cues.

Subtitles for the Deaf and Hard of Hearing

The standard cabling interface for carrying uncompressed video and embedded audio in facilities.

Serial Digital Interface

SDR (Signal-to-Distortion Ratio)

A standard metric for the quality of separated audio; higher SDR means cleaner separation.

Server-Side Ad Insertion

Stitching ads into the media stream at the server rather than the player.

SSAI; Ad Stitching

SIGMOS (Signal Mean Opinion Score)

A machine-estimated speech-quality score that predicts how good audio would sound to human listeners.

A representation of sound as it varies over time, as an electrical voltage or digital samples.

SI-SDR (Scale-Invariant Signal-to-Distortion Ratio)

A refinement of SDR that ignores volume scaling, giving a more reliable measure of separation quality.

An audio artifact where transients or frequencies blur together, reducing clarity.

A standard for carrying video, audio, and data as separate synchronized streams over IP in professional broadcast.

ST 2110; SMPTE 2110

A standardized clock that labels each video frame for synchronization.

SMPTE TC; Timecode

SNR (Signal-to-Noise Ratio)

A measure of how strong the wanted signal is relative to background noise; higher SNR is cleaner.

The individual sound-producing elements in a recording, such as a voice, instrument, or effect.

Audio that conveys a sense of direction and space, positioning sounds around the listener.

Speaker Diarization

Determining who spoke when in a recording by segmenting and labeling each speaker's turns.

A filter that selects which parts of a spectrogram belong to a given source, used to separate it from a mix.

A visual representation of audio showing how its frequency content changes over time.

Speech Enhancement

Improving the clarity and quality of speech in a recording by reducing noise and distortion.

Voice Enhancement

SRT (Secure Reliable Transport)

An open-source protocol for delivering broadcast-quality video and audio reliably over the public internet, with a configurable latency buffer.

SRT (SubRip Subtitle)

A plain-text subtitle file format with simple start/end timecodes.

An individual audio track isolated from a mix, such as vocals, drums, bass, or dialogue.

Audio with two channels (left and right) that creates a sense of width and placement.

STOI (Short-Time Objective Intelligibility)

An objective metric that predicts how intelligible speech is, on a 0-to-1 scale.

Licensing music to pair with visual media such as film, TV, ads, or games.

Sync Licensing; Synchronization

In source separation, the specific source you aim to isolate, especially an algorithm's estimate of it.

Target Speaker Extraction

Isolating one specific person's voice from a recording containing multiple speakers or noise.

A representation of audio as a waveform, showing how the signal's amplitude changes over time.

Waveform Domain

The maximum level of the reconstructed analog waveform, used to prevent clipping.

True Peak Level

An XML-based timed text standard used for subtitles and captions.

Timed Text Markup Language

The natural back-and-forth exchange of speaking turns in a conversation, human or with a voice agent.

Converting audio with fewer channels (e.g., mono or stereo) into a format with more, like 5.1 or Atmos.

A single continuous unit of speech, such as a spoken phrase or sentence, bounded by pauses.

VAD (Voice Activity Detection)

Technology that detects when human speech is present in an audio stream versus silence or noise.

Vocal Isolation

Extracting the vocals from a song, separated cleanly from the instrumental backing.

Voice Isolation; Vocal Remover; Vocal Splitter

The delay between a person finishing speaking and an AI voice agent responding.

Response Latency; Turn Latency

The most common uncompressed, lossless audio format; full quality but no embedded metadata.

The web-native caption format used by HTML5 video and HLS streaming.

VTT; Web Video Text Tracks

WER (Word Error Rate)

A standard metric for transcription accuracy; the share of words transcribed incorrectly, where lower is better.

An audio signal with effects or reverb applied; the opposite of dry.

Compact numerical 'fingerprints' of a speaker's voice, used to tell speakers apart.

Speaker Embeddings

An AI model's ability to handle a task, language, or input it was never explicitly trained on.

A commercial protocol and platform for reliable live video transport over IP.

No items found.

No terms match that search. Try a shorter word, or browse the A–Z list.