Free English accent checker

Accent Test

Read this passage aloud

About 30 seconds · normal speaking pace
Every Thursday, my brother Victor rides thirty miles to the village market. He usually buys three fresh loaves of bread, a bottle of olive oil, and five heavy bags of rice. Last month the weather was worse than we thought, so he asked a friend with a van to help. They laughed about it later, sitting by the window with hot tea, watching the rain fall on the little red roofs.

Your recording is sent to our AI model provider for analysis and is not stored. 5 free tests a day, 20 when signed in.

Read a 30-second passage aloud and find out which accent your English sounds closest to. The test names the two nearest accents, points to the exact words that give it away, and explains how to make each sound — plus whether you lean American or British. Free, no sign-up, and your recording is not stored.

Free, no sign-upWord-level feedbackRecording not stored

How it works

Three steps, about a minute

  1. 1

    Read the passage aloud

    Press record and read the passage at your normal pace — about 30 seconds. No microphone? Upload a recording of yourself reading it.

  2. 2

    Get your accent profile

    In about ten seconds you see the two accents your English sounds closest to, the exact words that point there, and a tip for each sound.

  3. 3

    Hear it read by a native voice

    Send the passage to the generator below and compare your reading with an American or British AI voice, word for word.

Examples

Practice passages, read by native AI voices

Every clip below was generated here on the free tier — no account, no credits. Play one to hear the voice on that kind of writing, or load the script into the generator and make your own version of it.

The th sounds

Think, this, three — the sound most accents change first.

Think about the three things that matter this Thursday: the weather, the theatre tickets, and the thirty thank-you notes that my brother and father still have to write.

Read by Adam · Clear, confident

V and W

Very well, not "wery vell" — lips versus teeth.

We visited a very quiet village one wet Wednesday. The view from the window was wonderful, and we waved at every van that drove over the old wooden bridge.

Read by Emma · British, calm

Short and long vowels

Ship or sheep, full or fool — length changes the word.

Sit in this seat while I fill the cup until it is full. The ship leaves at six, so eat quickly, pull your bag from the pool, and please do not let the sheep sleep on the deck.

Read by Bella · Warm, natural narration

Included

What you get

Two closest accents, never a score
The model names the two accents your pronunciation sits nearest to. It does not grade you — a percentage would claim a precision the method does not have.
Evidence from your own words
Every finding points at a word you actually read — "three" heard as "tree", "village" as "willage" — not a generic list of what an accent usually does.
A tip for each sound
Each feature comes with one practical instruction for making the standard sound, so the result doubles as a short practice plan.
American or British lean
The passage is built to expose the three markers that separate the two, so native speakers learn something too.
Recording not stored
The audio is converted in your browser, analyzed once, and dropped. Nothing is saved to your account or used for training.
Free, no sign-up
5 tests a day as a guest and 20 with a free account. No app to install.

What this accent test listens for

An accent is mostly a handful of sounds your first language handles differently from English. The passage above is written so that one 30-second read puts each of those sounds in front of the model several times — three th words in the first sentence alone — which is why a short recording is enough.

The model compares how you said each word with how it is said natively, then matches the pattern of differences against the accents it knows. The table shows what it is listening to, and the shifts that are common for some language backgrounds. Your result reports only what it heard in your recording, not this list.

SoundWhere it is in the passageCommon shifts
th — think, thisThursday, thirty, three, thought · the, brother, weather, theyOften heard as t / d (for example Hindi, Italian, Russian speakers) or as s / z (French, German, Japanese, Mandarin).
v and wVictor, village, van, heavy · worse, with, window, watchingb for v (Spanish, Japanese, Korean), f for v (Arabic), v for w (German), or one sound between the two (Hindi).
r and lrides, rice, rain, red, roofs · little, laughed, later, fallTapped or rolled r (Spanish, Italian, Russian, Arabic), r made in the throat (French, German), r and l blurred (Japanese, Korean).
Short vs long vowelssitting vs three, tea · full vs roofsPairs like ship / sheep merge in languages with five-vowel systems (Spanish, Italian, Japanese, Russian).
Word endingsasked, month, bags, loaves, roofsFinal consonant clusters shortened (Mandarin, Vietnamese, Indonesian) or given an extra vowel (Japanese).

How accurate is it?

We tested the method on 30 real recordings from the Speech Accent Archive — five speakers each of Spanish, Russian, Mandarin, French, Arabic and Hindi, reading that archive’s own passage. It named the right first language for 27 of the 30. Two of the three misses were placed as Russian. We also run eight synthetic accented voices through the live version after every change, and it currently names the right first candidate for all eight.

That is why the result says your pronunciation is closest to an accent, and never claims to know your first language: at nine in ten, the first sentence is honest and the second would be wrong one time in ten. It is also why there is no score. In testing, the model was exactly as confident on its wrong answers as on its right ones, so a percentage would look precise without meaning anything.

Accuracy drops with background noise, very short readings, speakers who learned English early or grew up with two languages, and languages the model has heard less of. Treat the result as a well-informed second opinion — not a certificate, and never a basis for hiring or admissions decisions.

American or British? What the passage reveals

Three markers separate the two main standards, and the passage contains each of them several times. If you are a native speaker, this is the part of the result worth reading: it tells you which way your English leans and what to change if you want to sound more like the other.

MarkerAmericanBritish
r after a vowel — worse, Thursday, marketPronouncedDropped
The vowel in last, laughed, askedShort, as in "cat"Long, as in "father"
The t in bottle, little, laterA quick d-like flapA clear t, or a glottal stop

An accent is not a mistake

Everyone has an accent, native speakers included. What matters for being understood is a small number of sounds that change meaning — ship and sheep, very and berry — not whether you sound like a newsreader. Listeners adapt to a steady accent within minutes; they struggle with sounds that turn one word into another.

So read the result as a map, not a grade. If you want to work on it, pick the one feature in your result that changes a word’s meaning, practise it with the drills above for a week, then take the test again and see whether it disappears from the list.

Hear the passage read by a native voice

The fastest way to hear the difference is side by side. The generator below is loaded with the same passage — play it in an American voice, switch to Emma for British English, and slow it down to hear exactly where each sound falls.

Speech studio

32 voices · 18 languages

5,000 free characters/day
376 / 1,000
Pick a voicePlay any of them
Free voices
Natural voices3 cr / 1k chars
Premium voices10 cr / 1k chars
Ultra voices20 cr / 1k chars

What happens to your recording

Your browser converts the recording to a small mono audio file and sends it once to our AI model provider, which returns the analysis. We do not store the audio, attach it to your account, or use it to train anything. The only thing we keep is a daily count of tests per visitor, so the free limit works.

Questions

Frequently asked