Pronounce Any English Word in 6 English Accents

Ready to test your ear?

Hearing a word is the easy half. Telling two near-identical words apart — ship and sheep, bed and bad — is what actually moves your pronunciation, and it is the part a lookup tool cannot do for you.

Quiz yourself Compare two accents

The same word sounds different across English accents. Hear each one, or play two back to back on Compare.

Accents, stress and phonetics

Three things decide whether an English word sounds right, and only one of them is about individual sounds. The first is the phoneme inventory: which contrasts your accent makes at all. English distinguishes the vowels in ship and sheep; many languages do not, and a speaker whose first language does not make that contrast will not hear it before they can produce it.

The second is word stress. English is a stress-timed language: one syllable in every content word carries the beat, and the rest are compressed around it. Put the stress in the wrong place and a word can become unintelligible even when every individual sound is correct — hotel is ho-TEL, not HO-tel, and record changes meaning depending on which half you lean on. In IPA the stressed syllable is the one after the ˈ mark, which is why our word pages show the transcription rather than just playing audio at you.

The third is reduction. Unstressed syllables in English collapse towards the schwa, /ə/ — the vowel at the start of about and the end of sofa. It is the most common vowel in the language and it is almost never written down, which is why English spelling is such an unreliable guide to English speech. Learning to reduce is usually what separates "correct" from "natural".

Start with the IPA guide if the symbols are new, or go straight to the full guide index.

Six accents are offered up front. They are not six labels on the same voice — each is a different variety with its own contrasts.

Most pronunciation sites stream audio from a server. This one asks your device first. When you press Pronounce, the tool looks through the speech voices your operating system already has — the ones behind the browser's Web Speech API — and uses one if it can. That audio is instant, works with no connection, and never leaves your machine, so nothing you type is transmitted anywhere.

Only when your device has no voice for the accent you chose does the request go outward, to a neural engine (Deepgram Aura-2 on Cloudflare Workers AI) that returns an audio clip. There is a second, less obvious case: some browsers list voices that never actually produce sound — Chrome's network-dependent voices are the usual culprit — so if a device voice is selected and stays silent, the tool notices and re-plays the word through the server voice instead. That is the "only British works" failure mode, fixed by falling back on silence rather than trusting the list.

The fallback is deliberately narrow: it only ever swaps within the same accent. The server engine has American, British and Australian voices and nothing else, so Indian, Canadian and Irish are device-only. If your device cannot speak them, the tool tells you and offers to help you install the voice — it will not hand you a British recording under an Indian label. Speed is applied in the browser via playback rate, so slow, normal and fast are one clip at three rates rather than three separate synthesis calls.

You cannot reliably produce a sound you cannot hear. Infants start out able to discriminate the phonetic contrasts of every human language and lose the ones their environment does not use — a process largely complete within the first year. By adulthood, a sound that is not contrastive in your first language is not merely hard to say; it is genuinely hard to hear, and it gets sorted into the nearest category you do have.

That is why pronunciation practice built only on repetition stalls. If /ɪ/ and /iː/ arrive in your ear as the same vowel, then ship and sheep sound identical, your own two attempts sound identical to you as well, and you have no signal telling you which one was right. Perception has to move before production can.

The standard training tool is the minimal pair: two words differing in exactly one sound, played until the contrast becomes audible. Forced-choice listening with immediate feedback is the format that works, because it makes you commit to an answer and then corrects you — which is precisely what our listening quiz does. Research on perceptual learning suggests focused listening practice is a more efficient use of time than repetition alone.

Browse the full word index, or type any other word above.

All 103 practice lists, each with audio.