AI accent identifier Identify your English accent.
Speak for 30 seconds. The identifier compares your sounds with 23 English accents and shows the closest matches with confidence scores.
Read this aloud in your normal voice (about 30 seconds):
Last summer, my brother and I took the early train to the coast. We bought three thick sandwiches, a bottle of water and a map, and we walked along the cliffs until the weather turned. By the time the rain arrived, we were laughing in a tiny cafe, drying our socks by the heater.
Free, no sign-upAudio never storedResult in under a minute
AccentLoop’s accent identifier analyses a 30-second recording of you reading a fixed passage and identifies the English accents it is closest to, out of 23, each with a confidence percentage. It also shows the evidence: the specific sounds, like a dropped r or a d-like t, that led to the answer. Free, no account, audio not stored.
How to identify your accent
- Press “Start the accent test” and allow the microphone. Nothing is recorded before you press it.
- Read the short passage on screen out loud at your normal pace — about 30 seconds.
- Press stop. In about 6 seconds you see the accents you sound closest to, with percentages.
- Check the US–UK scale, accent strength and the 2–3 sounds that gave your accent away.
- Share your result card or practise the exact sounds it found.
Sample voices and what the test said
Same passage, different speakers. Press play, then compare with the result.
Sample voice · Australian
Last summer, my brother and I took the early train to the coast. We bought three thick sandwiches, a bottle of water, and a map, and we walked along the cliffs until the weather turned. By the time the rain arrived, we were laughing in a tiny cafe, drying our socks by the heater.
Sample recorded with a synthetic voice reading our test passage, so you can hear the test before you take it.
Your result
You sound closest to Australian
What we heard
- Non-rhotic pronunciation (dropping 'r' at the end of 'summer', 'brother', 'water', 'weather', 'heater')
- Broad 'a' vowel in 'last' and 'laughing' (BATH vowel)
- Flapped 't' in 'water' combined with a raised vowel quality
- Shifted diphthongs in 'train', 'rain', and 'coast' characteristic of Southern Hemisphere English
An AI estimate for entertainment and practice — not a professional assessment. No accent is better or worse than another.
How this accent test compares
| AccentLoop | BoldVoice Accent Oracle | Fluently Accent Guru | Vocal Image accent test | |
|---|---|---|---|---|
| Recording length | ~30 s passage | ~30 s | ~30 s | 15 s (per its title) |
| Sign-up before the result | No | Not stated on page | No (per Fluently) | Not stated |
| Result | Top 3 closest accents with %, US–UK scale, strength, clarity, the sounds behind it | Accent guess + share image | Accent guess | “Worldwide Accent Score” + prediction |
| What happens next | Practise the exact sounds in the browser | App trial / QR code to download | App | App download badges |
| Hear sample results first | Yes — 14 sample voices below | No | No | No |
Competitor details are from their public pages, checked 18 Oct 2026. Features and prices change — tell us if something is out of date.
The clues an accent identifier uses
| Clue | Example | Points toward |
|---|---|---|
| Rhoticity (r after vowels) | “summer” with or without r | With r: US, Canada, Ireland, Scotland, India. Without: England, Australia, South Africa |
| Flap t | “water” → “wadder” | US, Canada, Australia |
| Trap–bath split | “laugh” with long ah | England, Australia, South Africa |
| th-stopping / fronting | “three” → “tree” or “sree” | Irish English; Spanish, German, Russian, Korean speakers |
| Syllable timing | every syllable the same length | Indian, Singaporean, Nigerian English; Mandarin, Spanish, Japanese speakers |
| Vowel insertion | “took” → “took-u” | Japanese, Korean, Brazilian Portuguese speakers |
How confident is the identification?
Every result has a percentage. One accent above about 70% means a clear match. Two accents close together usually means a genuinely mixed accent — common for people who moved countries or learned English from media. Low scores across the board mean the recording was too noisy or too short; try again in a quiet room.
On the 14 sample voices on this site (synthetic voices, each reading our passage in a different accent) the test’s first guess matched the intended accent 13 times out of 14. The Latin American Spanish sample came out as Spanish (Spain) 55%, Latin American 30%.
The 23 accents the test can name
| Group | Accents |
|---|---|
| North America | General American, Canadian |
| British Isles | Southern British (RP-like), Scottish, Irish |
| Southern hemisphere | Australian, New Zealand, South African |
| Asia & Africa (English as an official language) | Indian, Nigerian, Filipino, Singaporean |
| English with a first-language influence | Mandarin, Spanish (Spain), Latin American Spanish, French, German, Russian, Arabic, Japanese, Korean, Brazilian Portuguese, Vietnamese |
If your voice doesn’t fit any of them well, the result says so instead of forcing a label.
Frequently asked questions
Can AI identify accents?
Yes, reasonably well for clear recordings. It compares sounds such as r after vowels, t between vowels, vowel length and rhythm with patterns typical of each accent and gives a confidence for the closest ones. Mixed accents get lower, split scores.
Can I identify someone else’s accent?
Only with their consent and their own voice: ask them to take the 30-second test on their device. We don’t accept uploaded recordings of other people.
Is the accent test free?
Yes. The accent test is free with no account and no email. To stop abuse there is a soft limit of a few tests per day per connection.
Do you keep my recording?
No. The audio is analysed and then discarded — it is never stored. We keep the text result for 30 days so your share link works, and you can delete it any time from your account page. We never create a voiceprint.
How accurate is it?
Before launch we tested it on 10 synthetic voices in different native English accents and it named the right region 9 times out of 10; on the 14 sample voices on this site its first guess matched 13 of 14. Real voices, especially mixed and second-language accents, are harder, so treat the result as “closest to”, not a label. It is for entertainment and practice, not a professional assessment.
Is having an accent bad?
No. Everyone has an accent. The test only tells you which accents you sound closest to and which sounds create that impression, so you can decide what — if anything — you want to change.
Last updated 2026-10-18. Results are estimates for practice and entertainment, not a professional or official assessment.