Open reference

Listening words: how to describe a voice

Institute of Thinkinginstituteofthinking.comVersion 1 · 28 Oct 2026

Shared words for describing what we hear in a voice clip. Institute of Thinking made this list for its own research and shares it under the CC BY 4.0 licence. Each word has one plain meaning and, where it helps, an example sentence. A word with a note number is defined from that source; a word without one is our own working wording, written with the help of Claude, an AI assistant.

How to report what you hear

Say where, what, how strong: "At second twelve, the man's voice goes metallic, strong." Strength: slight, clear, strong. One clip, one note per problem.

Pitch and melody

Prosody is the group name for intonation, stress, rhythm and loudness: the music on top of the sounds ¹.

Word	What it means	Say it like
Pitch	How high or low the voice sits.	"Pitch too low for a young woman."
Pitch range	How far the voice moves between its high and low notes.	"Narrow range, almost one note."
Intonation	The melody of a sentence: where it rises and falls.	"Questions don't rise at the end."
Monotone	Flat melody, no ups and downs.	"Monotone, sounds like reading."
Sing-song	Melody that goes up and down too much, or always the same way.	"Sing-song, every line ends the same."
Rising end	Pitch goes up at the end, as in a question; on a statement it is uptalk ².	"Every line rises, sounds unsure."
Loudness	How loud a sound feels ³.	"Too quiet at the end of lines."

Tone: the colour of the voice

Tone (also timbre) is what makes two voices differ on the same note.

How the vocal folds make sound is phonation ⁴; the everyday one is modal voice ⁵.

Word	What it means
Warm	Soft, round, rich in the low part. Pleasant, close.
Bright	Sharp, clear, rich in the high part. Can turn harsh.
Dark	Deep, covered, a bit heavy.
Thin	Little body, sounds small or far away.
Full	Lots of body, sounds big and close.
Nasal	Too much air through the nose while speaking.
Breathy	Air leaks with the voice, like a whisper mixed in.
Creaky	Low rattle, like a slow door, often at line ends (also vocal fry).
Hoarse / raspy	Rough, scratchy, like a sore throat or a smoker.
Tense / pressed	Squeezed, strained, as if pushing.
Relaxed	Easy, loose, no effort heard.
Whisper	No vocal-fold buzz at all, only air ⁶.
Harsh	Squeezed throat, rough and strained; also pressed ⁷.
Falsetto	The register just above the normal voice, thin and high ⁸.
Formants	The resonances of the mouth and throat; they make an a sound like an a ⁹.

Rhythm and timing

Word	What it means	Say it like
Rhythm	The beat of speech: how long and short parts follow each other.	"Rhythm is even, like a robot."
Pace	How fast someone speaks.	"Pace too fast, words run together."
Pause	A silence inside or between sentences.	"Pause in the wrong place, mid-word."
Stress	The part of a word said louder or longer (CAsa, not caSA) ¹⁰.	"Stress on the wrong syllable."
Flow	Speech moves on without stumbling.	"Flow breaks, it hesitates."
Choppy	Cut into small pieces, no flow.	"Choppy, like words glued together."
Drawl	Sounds pulled long and slow.	"Vowels drawled, very southern."
Clipped	Sounds cut short and quick.	"Clipped endings, swallowed vowels."
Disfluency	A break in fluent speech: false starts, repeats, "ehm" ¹¹.	"Lots of false starts."
Filler	A sound or word ("ehm", "cioè") that holds the turn while thinking ¹².	"No fillers, too clean."
Speaking vs articulation rate	Speed with the pauses counted, and without them ¹³.	"Talks fast but pauses a lot."

Conversation

Word	What it means	Say it like
Turn	One person's go at speaking.	"Its turns are too long."
Gap	Silence between one person stopping and the other starting.	"Gap too long, feels slow."
Overlap	Both speak at the same time.	"Overlap at the end of my line."
Interruption	One cuts in before the other has finished.	"It interrupts me mid-sentence."
Barge-in	I cut in and the system should stop and listen.	"Barge-in fails, it keeps talking."
Backchannel	Small listening sounds: mhm, sì, ah.	"No backchannel, feels cold."
Talking over	Keeps speaking through the other person.	"It talks over me."
Turn-taking	Speaking one at a time, in turns ¹⁴.	"Turn-taking feels natural."
IPU	A stretch of speech with at least 0.2 s of silence each side; Moshi measures pauses, gaps and overlaps with it ¹⁵.	—

Accent and pronunciation

Word	What it means	Say it like
Accent	The way a place or group colours speech: melody, vowels, rhythm.	"Neapolitan melody, but northern vowels."
Pronunciation	How a word's sounds are made.	"Gli sounds like li."
Vowel colour	Whether a vowel is open or closed (è vs é).	"Open e where Rome says closed."
Double consonants	Held consonants in Italian (fatto, not fato).	"Doubles missing, sounds foreign."
Mumbled	Sounds not finished, mouth barely moves.	"Mumbled ending."
Slurred	Sounds melt into each other, like drunk.	"Slurred, hard to follow."
Foreign tinge	Sounds like a non-Italian speaking Italian.	"Slight English tinge on the r."
Articulation	How the tongue, lips and palate shape each sound ¹⁶.	"Sloppy articulation."
Intelligibility	How easy the speech is to understand ¹⁷.	"Hard to understand in the noise."
Coarticulation	A sound changing to be more like its neighbour ¹⁸.	—
Elision	A sound or syllable left out ('sto for questo) ¹⁹.	"Too much elision, sounds lazy."
Code-switching	Changing language or dialect inside a conversation ²⁰.	"Switches to English mid-sentence."

Feeling and delivery

Word	What it means
Expressive	Feeling comes through: happy, tired, annoyed.
Flat	No feeling, neutral to the point of dead.
Energy	How awake and pushed the voice is.
Emphasis	Some words pushed to carry meaning.
Read, not spoken	Sounds like reading a text, not talking.
Natural	Could pass for a real person in a real moment.
Paralanguage	What the voice adds beyond the words: pitch, volume, tone, feeling ²¹.
Laughter	Rhythmic breath bursts, usually heard as joy or relief ²²; real talk has laughs, sighs and coughs.

Noises from the recording

Word	What it sounds like
Background noise	Anything behind the voice: street, TV, people.
Hiss	Steady ssss, like air or old tape.
Hum / buzz	Steady low tone, like a fridge or a bad cable.
Echo	Clear repeats of the voice, like a big empty hall.
Room sound	Voice bounces off walls, sounds like a bathroom (also reverb).
Muffled	Like talking through a blanket; the highs are gone.
Tinny	Like a cheap speaker or a tin can; the lows are gone.
Phone sound	Narrow, boxy, like an old phone line.
Clipping	Too loud, cracks and distorts on loud parts.
Pops	Bursts of air on p and b.
Mouth noise	Lip smacks, tongue clicks, wet sounds.
Breaths	Audible in-breaths. Some are natural; too many is not.
Pumping	Background rises and drops as the voice starts and stops.

Noises made by the AI voice

Word	What it sounds like
Metallic / robotic	Buzzy, ringing, like a voice through a metal pipe.
Warble	Wobbly, bubbly, like underwater.
Glitch	A short jump, stutter or broken bit.
Cut-off	Word or sentence stops before the end.
Click	A tiny tick or snap where pieces join.
Drift	The voice slowly turns into another person.
Accent drift	The accent fades or changes during the clip.
Babble	Speech-like sounds with no real words.
Wrong words	Says something not in the text, skips or repeats words.
Wrong speed	Sounds sped up or slowed down, chipmunk or sleepy.
Silence holes	Unexpected dead air inside speech.
Repetition	The same word or phrase over and over.
Wrong language	The recogniser hears another language.
Low naturalness	A program rates the clip as unnatural.
Compression artifact	Damage left by squeezing audio too hard ²³.

Words used to test the model

Word	What it means
Full-duplex	Both sides can talk at once, like a phone call ²⁴; Moshi listens while it speaks ²⁵.
Half-duplex	Turn by turn: these systems "do not listen while speaking" ²⁶.
Latency	Seconds from the end of the user's speech to the start of the answer ²⁷.
Takeover (TOR)	The model takes the turn instead of staying quiet or backchannelling ²⁸.
Pause handling	Not jumping in when the user only pauses ²⁹.
Smooth turn-taking	Answering promptly when the user ends, without long delay or overlap ³⁰.
User interruption	The user cuts in; the model should notice and adapt ³¹.
Voice activity detection (VAD)	A program deciding when someone is speaking ³².
Voice prompt	A short speech sample that tells PersonaPlex which voice to use ³³.
Text prompt (role)	Text that tells PersonaPlex its role and scenario ³⁴.
Inner monologue	Moshi writes the text of its own words just before saying them ³⁵.
Codec	A program that squeezes audio into a few numbers and back ³⁶; Moshi's is Mimi.
MOS	Mean opinion score: the average of people's quality ratings ³⁷.
WER (wrong words)	Share of words wrong; 0 means identical ³⁸.
Voice match	How sure a program is that two clips are the same person ³⁹.
Emotion recognition	A program naming the feeling in a voice, face or text ⁴⁰.

Sounds and words: the basics

A phoneme is the smallest sound that tells one word from another, written between slashes ⁴¹.
A vowel is said with the mouth open and little or no blocking; a consonant closes the way fully or in part ⁴² ⁴³.
A syllable is built around a nucleus, most often a vowel ⁴⁴. Stress is the extra weight on one syllable: louder, longer, or a change of pitch ⁴⁵.
A consonant is described by voice (do the vocal folds vibrate) ⁴⁶, place (where the narrowing is) ⁴⁷ and manner (how the air passes) ⁴⁸.
The manners: stop blocks the air completely ⁴⁹; fricative forces it through a narrow gap ⁵⁰; affricate starts as a stop and ends as a fricative ⁵¹; nasal lets the air out through the nose ⁵²; liquid covers the l- and r-sounds ⁵³.
A double consonant (gemination) is a consonant held longer ⁵⁴; in Italian fato and fatto are two words ⁵⁵. A diphthong is two vowel sounds in one syllable; split across two syllables they are a hiatus ⁵⁶. Italian ai and au are common ⁵⁷.
The places of the Italian consonants follow the table of Italian phonology: lips, teeth and gum ridge, palate, soft palate ⁵⁸.

Speech sounds: phoneme, syllable, word; vowel and consonant; voice, place and manner; stress, double consonant, diphthong

The phonetic alphabet (IPA) for Italian

The IPA (International Phonetic Alphabet) is one standard set of symbols for the sounds of speech, used in every language ⁵⁹.
There is no separate alphabet for Italian. Italian uses a part of the same IPA; Wikipedia's Italian key lists the symbols, the example words and the marks shown above ⁶⁰.
One Italian linguist, Luciano Canepari, built an extended version called canIPA, for finer detail; it has seen little use outside his own work ⁶¹.
Italian has seven vowels in stressed syllables; the open ɛ ɔ appear only there, the closed e o elsewhere ⁶². Palermo pronounces them all open and parts of Piedmont in between, so this is where accents show ⁶³.
The letters s and z each hide two sounds: s is [s] or [z], z is [ts] or [dz] ⁶⁴.

The Italian vowels on the vowel chart, the consonants whose IPA symbol differs from the letter, and the IPA marks for stress, length and syllable break

Licence and how to cite

Text and infographics: CC BY 4.0, Institute of Thinking, 2026. You may copy, share and adapt them with credit. Cite as: Institute of Thinking, Listening words: how to describe a voice, version 1, 28 October 2026, instituteofthinking.com. Wording taken from a source in the notes keeps that source's own licence.

Sources

Wikipedia contributors (2026). Prosody (linguistics). https://en.wikipedia.org/wiki/Prosody_(linguistics). p. 1, lead section, revision 1377877018.
Wikipedia contributors (2026). High rising terminal. https://en.wikipedia.org/wiki/High_rising_terminal. p. 1, lead section, revision 1372093522.
Wikipedia contributors (2026). Loudness. https://en.wikipedia.org/wiki/Loudness. p. 1, lead section, revision 1381124040.
Wikipedia contributors (2026). Phonation. https://en.wikipedia.org/wiki/Phonation. p. 1, lead section, revision 1378319920.
Wikipedia contributors (2026). Modal voice. https://en.wikipedia.org/wiki/Modal_voice. p. 1, lead section, revision 1364125584.
Wikipedia contributors (2026). Whispering. https://en.wikipedia.org/wiki/Whispering. p. 1, lead section, revision 1359390609.
Wikipedia contributors (2026). Harsh voice. https://en.wikipedia.org/wiki/Harsh_voice. p. 1, lead section, revision 1332671154.
Wikipedia contributors (2026). Falsetto. https://en.wikipedia.org/wiki/Falsetto. p. 1, lead section, revision 1377132866.
Wikipedia contributors (2026). Formant. https://en.wikipedia.org/wiki/Formant. p. 1, lead section, revision 1340709526.
Wiktionary contributors (2026). casa. https://en.wiktionary.org/wiki/casa#Italian. p. 1, Italian, Pronunciation, revision 93683508.
Wikipedia contributors (2026). Speech disfluency. https://en.wikipedia.org/wiki/Speech_disfluency. p. 1, lead section, revision 1369279759.
Wikipedia contributors (2026). Filler (linguistics). https://en.wikipedia.org/wiki/Filler_(linguistics). p. 1, lead section, revision 1382079903.
Wikipedia contributors (2026). Speech tempo. https://en.wikipedia.org/wiki/Speech_tempo. p. 1, paragraph on pauses and hesitations, revision 1322959091.
Wikipedia contributors (2026). Turn-taking. https://en.wikipedia.org/wiki/Turn-taking. p. 1, lead section, revision 1368892558.
Alexandre Défossez, Laurent Mazaré, Manu Orsini et al. (2024). Moshi: a speech-text foundation model for real-time dialogue. https://arxiv.org/abs/2410.00037. §5.6.p1 "Metrics", p. 31.
Wikipedia contributors (2026). Manner of articulation. https://en.wikipedia.org/wiki/Manner_of_articulation. p. 1, lead section, revision 1363377714.
Wikipedia contributors (2026). Intelligibility (communication). https://en.wikipedia.org/wiki/Intelligibility_(communication). p. 1, lead section, revision 1374851940.
Wikipedia contributors (2026). Coarticulation. https://en.wikipedia.org/wiki/Coarticulation. p. 1, lead section, revision 1341975227.
Wikipedia contributors (2026). Elision. https://en.wikipedia.org/wiki/Elision. p. 1, lead section, revision 1367520909.
Wikipedia contributors (2026). Code-switching. https://en.wikipedia.org/wiki/Code-switching. p. 1, lead section, revision 1380977755.
Wikipedia contributors (2026). Paralanguage. https://en.wikipedia.org/wiki/Paralanguage. p. 1, lead section, revision 1379872409.
Wikipedia contributors (2026). Laughter. https://en.wikipedia.org/wiki/Laughter. p. 1, lead section, revision 1377416505.
Wikipedia contributors (2026). Compression artifact. https://en.wikipedia.org/wiki/Compression_artifact. p. 1, lead section, revision 1382098074.
Wikipedia contributors (2026). Duplex (telecommunications). https://en.wikipedia.org/wiki/Duplex_(telecommunications). p. 1, lead section, revision 1379236825.
Défossez et al., §1 "Introduction", pp. 2–3.
Rajarshi Roy, Jonathan Raiman, Sang-gil Lee et al. (2026). PersonaPlex: Voice and Role Control for Full Duplex Conversational Speech Models. https://arxiv.org/abs/2602.06053. §2 "Related Work", pp. 1–2.
Guan-Ting Lin, Jiachen Lian, Tingle Li et al. (2025). Full-Duplex-Bench: A Benchmark to Evaluate Full-Duplex Spoken Dialogue Models on Turn-taking Capabilities. https://arxiv.org/abs/2503.04721. pp. 3–5, section III.B, Evaluation Dimensions.
Lin et al., §III "Full-Duplex-Bench Framework", p. 3.
Lin et al., pp. 3–5, section III.B, Evaluation Dimensions.
Lin et al., pp. 3–5, section III.B, Evaluation Dimensions.
Lin et al., pp. 3–5, section III.B, Evaluation Dimensions.
Wikipedia contributors (2026). Voice activity detection. https://en.wikipedia.org/wiki/Voice_activity_detection. p. 1, lead section, revision 1310320123.
Roy et al., §3.1 "Architecture", p. 2.
Roy et al., §3.1 "Architecture", p. 2.
Défossez et al., §3.4.4.p3 "Joint sequence modeling for Moshi", p. 17.
Wikipedia contributors (2026). Audio codec. https://en.wikipedia.org/wiki/Audio_codec. p. 1, lead section, revision 1354433960.
Wikipedia contributors (2026). Mean opinion score. https://en.wikipedia.org/wiki/Mean_opinion_score. p. 1, lead section, revision 1380628996.
Wikipedia contributors (2026). Word error rate. https://en.wikipedia.org/wiki/Word_error_rate. p. 1, lead section, revision 1281062858.
Wikipedia contributors (2026). Speaker recognition. https://en.wikipedia.org/wiki/Speaker_recognition. p. 1, lead section, revision 1380225450.
Wikipedia contributors (2026). Emotion recognition. https://en.wikipedia.org/wiki/Emotion_recognition. p. 1, lead section, revision 1380996887.
Wikipedia contributors (2026). Phoneme. https://en.wikipedia.org/wiki/Phoneme. p. 1, lead section, revision 1381374426.
Wikipedia contributors (2026). Vowel. https://en.wikipedia.org/wiki/Vowel. p. 1, lead section, revision 1380115729.
Wikipedia contributors (2026). Consonant. https://en.wikipedia.org/wiki/Consonant. p. 1, lead section, revision 1374049932.
Wikipedia contributors (2026). Syllable. https://en.wikipedia.org/wiki/Syllable. p. 1, lead section, revision 1380139128.
Wikipedia contributors (2026). Stress (linguistics). https://en.wikipedia.org/wiki/Stress_(linguistics). p. 1, lead section, revision 1381708585.
Wikipedia contributors (2026). Voice (phonetics). https://en.wikipedia.org/wiki/Voice_(phonetics). p. 1, lead section, revision 1373877429.
Wikipedia contributors (2026). Place of articulation. https://en.wikipedia.org/wiki/Place_of_articulation. p. 1, lead section, revision 1374908094.
Wikipedia, “Manner of articulation”, p. 1, lead section, revision 1363377714.
Wikipedia contributors (2026). Plosive. https://en.wikipedia.org/wiki/Plosive. p. 1, lead section, revision 1363377527.
Wikipedia contributors (2026). Fricative. https://en.wikipedia.org/wiki/Fricative. p. 1, lead section, revision 1382053993.
Wikipedia contributors (2026). Affricate. https://en.wikipedia.org/wiki/Affricate. p. 1, lead section, revision 1377342124.
Wikipedia contributors (2026). Nasal consonant. https://en.wikipedia.org/wiki/Nasal_consonant. p. 1, lead section, revision 1381336950.
Wikipedia contributors (2026). Liquid consonant. https://en.wikipedia.org/wiki/Liquid_consonant. p. 1, lead section, revision 1365850882.
Wikipedia contributors (2026). Gemination. https://en.wikipedia.org/wiki/Gemination. p. 1, lead section, revision 1376129427.
Wikipedia contributors (2026). Italian phonology. https://en.wikipedia.org/wiki/Italian_phonology. p. 1, section Consonants, revision 1374139897.
Wikipedia contributors (2026). Diphthong. https://en.wikipedia.org/wiki/Diphthong. p. 1, lead section, revision 1381622438.
Wikipedia, “Italian phonology”, p. 1, section Vowels, revision 1374139897.
Wikipedia, “Italian phonology”, p. 1, section Consonants, table of consonant phonemes, revision 1374139897.
Wikipedia contributors (2026). International Phonetic Alphabet. https://en.wikipedia.org/wiki/International_Phonetic_Alphabet. p. 1, lead section, revision 1382236090.
Wikipedia contributors (2026). Help:IPA/Italian. https://en.wikipedia.org/wiki/Help:IPA/Italian. p. 1, consonant, vowel and suprasegmental tables, revision 1368258444.
Wikipedia contributors (2026). Luciano Canepari. https://en.wikipedia.org/wiki/Luciano_Canepari. p. 1, lead section, revision 1342384212.
Wikipedia, “Help:IPA/Italian”, p. 1, note to the vowel table, revision 1368258444.
Wikipedia, “Italian phonology”, p. 1, section Vowels, revision 1374139897.
Wikipedia, “Help:IPA/Italian”, p. 1, consonant table and its notes, revision 1368258444.