4.7850+ ratings50K+ learners

Free · 6 sentences · about 2 minutes · no signup

AI Accent Detector: which sounds give your English accent away?

Read 6 sentences aloud. Free, no signup.

I think those three things are worth it.

th: think, those, three, worth

Tap to record

Every sound is scored 0–100 against a standard American English reference. The detector does not score intonation, sentence stress or fluency, and it does not guess where you are from.

This free AI accent detector listens to six short English sentences and tells you two things: how strong your accent is, as a 0–100 score for how close your speech comes to a standard American English reference, and which sounds give it away, from th, r and l to vowel pairs and word endings.

It does not guess your country or your native language from your voice, and it tells you why below. If you like, pick your first language: the detector then shows the pattern typical for speakers of that language next to your own results, so you can see which of the usual giveaways are really yours.

How the AI accent detector works

  1. 1

    Read six sentences

    Each sentence packs in the sounds listeners notice first: the two th sounds, r and l, v and w, b, f and p, z and j, sh, h, the vowel pairs of ship / sheep, full / fool and cat / bed, and word endings.

  2. 2

    Every sound is scored

    The English engine compares each sound in each word with a standard American English reference and gives it a score from 0 to 100. You see your word scores after every sentence.

  3. 3

    Accent strength

    After the sixth sentence the detector averages your sentence scores into one number and a band, from “very light” to “strong”. It is a measure of closeness to the reference, not of how well people understand you.

  4. 4

    The sounds that give you away

    Every sound group is scored by the median of all its occurrences across the six sentences, so one slip does not count against you. Groups below 80 are listed as your giveaways, lowest first, with how to fix each one.

Why this accent detector does not guess where you are from

Accent guessers such as BoldVoice’s Accent Oracle and AccentGuess answer “where is my accent from?”. That is a fun question, but it is not the useful one, and from a short recording it is a guess: an accent is a handful of specific sounds, and the same slip, th said as s or “ship” said as “sheep”, is made by speakers of many different first languages. We tried building a first-language guesser on our own engine in October 2026; it did not pass our accuracy check, so we did not ship it.

What a recording can tell you reliably is which sounds came out far from the reference. That is also the part you can act on: drill those sounds and your accent gets lighter, whatever your first language is. So this detector reports how strong your accent is and which sounds cause it, and leaves the guessing out.

What the accent strength score means

Your accent strength is the average of your six sentence scores. 90 and up is a very light accent, 80–89 light, 65–79 noticeable, and below 65 strong. These bands are this page’s own. They are not a CEFR level or an exam grade, and a strong accent can still be perfectly easy to understand.

As a check, on 30 October 2026 we read the six sentences with a synthetic American voice and a synthetic British voice: they averaged about 97 and 95. Real voices vary more than computer voices, so read your result as a direction, not a verdict.

American, British or another English?

The reference is standard American English, and the detector does not tell you which English accent you have. Most of the sounds it reports, th, r and l, v and w, the vowel pairs and word endings, are the same in American and British English. One is not: standard British English drops the r after a vowel (“park”, “for”, “are”), so a British speaker can see r flagged here. That is a difference from the American reference, not a mistake.

The sounds it checks

Fifteen groups, each heard at least three times across the six sentences. A group counts as a giveaway when its median score is under 80.

Sound groupHeard inHow to make it
th (think, this)think, three, those, withPut the tongue tip against the bottom edge of the top teeth and let the air flow over it; the tongue never taps. “this” adds voice, “think” does not.
r (red, girl)really, red, road, girl, worthCurl the tongue back or bunch it up so it touches nothing, and round the lips a little. No tap and no roll.
l (little, long)little, long, liked, full, schoolTouch the ridge just behind the top teeth with the tongue tip and keep it there through the sound, also at the end of “full” and “school”.
v (very, five)visited, very, five, ofTop teeth on the lower lip, voice on. The lips never close (that is b) and never round (that is w).
w (we, with)we, with, will, was, worthRound the lips as for “oo” and open them into the vowel; no teeth on the lip.
b (busy, black)busy, books, blackClose both lips fully, then release them with voice.
f (five, food)five, friends, full, foodTop teeth on the lower lip and blow; the lips never meet (that is p).
p (park, shop)park, happy, shopClose the lips and release them with a small puff of air, no voice.
z and j (busy, jacket)those, busy, his, was, jacketz is an s with the voice buzzing; j (jacket) starts with the tongue on the ridge, like d, and opens into a buzzing sh.
sh and s (she, sister)she, shop, sister, seatFor sh, round the lips and pull the tongue slightly back; for s, spread the lips and keep the tongue tip near the top teeth.
h (he, his)he, his, happyA breath from the throat before the vowel, like fogging a mirror. Do not drop it, and do not make it rough.
ship and sheep vowelssit / seat, it / he“sheep” (i) is long with spread lips; “ship” (ɪ) is short and relaxed, the jaw a little lower. Practise sit / seat back to back.
full and fool vowelsfull, books / food, school“fool” (u) is long with tightly rounded lips; “full” (ʊ) is short with relaxed lips.
cat and bed vowelsblack, jacket, asked / red, saidFor “cat” (æ) drop the jaw and spread the lips; for “bed” (ɛ) the mouth is half as open.
word endingsliked, asked, friends, old, booksFinish every final consonant: release the t of “liked”, keep both sounds of “-nds”, and add no extra vowel after it.

Typical giveaways by first language

Your first language shapes which English sounds are hard, because you hear and say English through its sound system. These are common tendencies, not rules: plenty of speakers have none of them. Pick your language above the recorder and the detector sets this pattern next to your own scores.

Spanish speakers
Spanish has five vowels, so ship/sheep and full/fool often merge, and unstressed vowels may stay full instead of becoming the schwa. Many speakers make b and v alike, add an e before s + consonant (“espeak”), and simplify final clusters.
Checked here: ship and sheep vowels, full and fool vowels, v (very, five), b (busy, black), word endings
Portuguese speakers
Brazilian speakers often add an “i” sound after a final consonant (“Facebooki”), say t and d before an i sound like ch and j, and replace th with t, f or s. An English r at the start of a word can sound like an h, final l can become w, and vowels before m or n are often nasalised.
Checked here: word endings, th (think, this), r (red, girl), h (he, his), l (little, long)
Russian speakers
Final voiced consonants are often devoiced (bag sounds like back), cat and bet can merge, w may come out as v, and th as s, z, t or d. Russian intonation patterns, carried over, can make English questions and statements sound flatter or more abrupt than intended.
Checked here: word endings, cat and bed vowels, w (we, with), v (very, five), th (think, this)
Chinese (Mandarin) speakers
Mandarin syllables rarely end in consonants, so final consonants may be dropped or followed by an extra vowel, and clusters simplified. Common points are th (often s or f), v and w, the l–n distinction for some speakers from southern China, vowel length, and an even, syllable-by-syllable rhythm with few reduced vowels.
Checked here: word endings, th (think, this), v (very, five), w (we, with), l (little, long), ship and sheep vowels
Japanese speakers
Japanese has one liquid where English has r and l, and words built from consonant + vowel beats, so a vowel is often added after consonants (“strike” said with extra syllables). th often becomes s or z, v becomes b, and the even, mora-timed rhythm leaves little room for the schwa.
Checked here: r (red, girl), l (little, long), word endings, th (think, this), v (very, five)
Korean speakers
Korean does not separate r and l the way English does, and has no f, v, z or th, so they often come out as p, b, j and s or d. Final stops are usually unreleased, and a vowel is sometimes added after a final consonant.
Checked here: r (red, girl), l (little, long), f (five, food), v (very, five), z and j (busy, jacket), th (think, this), word endings
Vietnamese speakers
Vietnamese words are single syllables with few possible final consonants, so English endings (especially -s, -ed and clusters) are often dropped. Speakers may carry tone-like pitch into English words and give each syllable similar weight, which hides the stress pattern.
Checked here: word endings, th (think, this), sh and s (she, sister)
Hindi speakers
Hindi has one sound between v and w, so the two often merge. English t and d may be said with the tongue curled back (retroflex), and th as a dental t or d. Many speakers stress different syllables from British or American English and give unstressed vowels their full value.
Checked here: v (very, five), w (we, with), th (think, this), r (red, girl)
Arabic speakers
Arabic has no p, so p and b can merge (park/bark), and v may come out as f. Many varieties have fewer vowels, so ship/sheep and bed/bid can merge, and a short vowel may be inserted to break up consonant clusters.
Checked here: p (park, shop), b (busy, black), v (very, five), f (five, food), ship and sheep vowels, cat and bed vowels
German speakers
w often comes out as v (“wery”), and final voiced consonants are devoiced, so “dog” ends like “dock”. th is often said as s or z.
Checked here: w (we, with), v (very, five), word endings, th (think, this)
French speakers
The h is often dropped (“is” for “his”) and th comes out as s or z. French stresses the last syllable of a word group, which English does not.
Checked here: h (he, his), th (think, this), ship and sheep vowels
Indonesian speakers
Indonesian has no th, and f, v, z and sh appear only in borrowed words, so they are often said as p, f or s. Final consonant clusters (asked, texts, world) are rare, so one sound tends to drop.
Checked here: th (think, this), f (five, food), v (very, five), z and j (busy, jacket), sh and s (she, sister), word endings

More free English accent tools

Questions about the AI accent detector

Can this AI accent detector guess where I’m from?+

No, on purpose. An accent is made of specific sounds, and the same sound slip is common to speakers of many first languages, so a guess of your country or native language from a short recording is often wrong. Our own test of a first-language guesser did not pass our accuracy check. Instead, the detector measures how strong your accent is and names the sounds that give it away. You can tell it your first language yourself to see the pattern typical for it.

How does the accent detector work?+

You read six short sentences into your microphone. The English engine scores every sound in every word against a standard American English reference. After the last sentence you get an accent strength score (the average of your sentence scores) and your sound groups ranked weakest first; groups with a median below 80 are listed as the ones that give your accent away.

Is the accent detector free?+

Yes. It runs in your browser with no signup and no download, up to 30 recordings a day. The TonePerfect app, on iPhone, Android and the web, adds daily practice for the sounds it finds.

What does my accent strength score mean?+

It is how close your speech came to the American English reference, averaged over six sentences: 90 and up is a very light accent, 80–89 light, 65–79 noticeable, below 65 strong. It is not a CEFR level, an exam grade or a measure of how easily people understand you.

Can it tell if I have an American or British accent?+

No. It scores you against one reference, standard American English, and does not classify English varieties. Most of the sounds it checks are the same in British English; the r after a vowel is not, so British speakers may see r flagged.

Does it score intonation, rhythm or fluency?+

No. It scores sounds. Intonation, sentence stress, rhythm and fluency are part of an accent too, but this detector does not measure them.

What is the difference between this and the English accent test?+

The English accent test is five single words and a quick ranking. The accent detector reads six full sentences, scores every sound in them, groups the sounds and gives one accent strength score, with your first language’s typical pattern if you choose it.

First-language tendencies from TonePerfect’s English guides; scores from TonePerfect’s own English speech engine.

Fix the sounds that give you away

TonePerfect has English lessons and daily speaking practice with a score for every sound, on iPhone, Android and the web. Free to start.