English to phonetic script
Paste English, get it in the phonetic alphabet in the two spellings dictionaries use, then hear it read by someone who grew up with the accent.
English spelling stopped tracking English pronunciation around five hundred years ago and never caught up. Choir is said “kwire”. Colonel is said “kernel”. Yacht has three letters that do nothing. There is no rule you can learn that fixes this, because there was no rule when the spellings were fixed.
The phonetic alphabet writes one symbol for each sound, so nothing is left to guess. This page writes your text in it twice: once the way Cambridge and Oxford write British pronunciation, once the way American dictionaries write American pronunciation. Then it reads the text aloud in a voice from a place where people speak that way.
Press the button to see the transcription.
What we change, and why
The phonetic script comes from eSpeak NG. eSpeak knows how to say English; it does not write English the way a dictionary writes it. Eight differences are corrected, and every one of them is listed here rather than left for you to notice.
| eSpeak writes | We write | Because |
|---|---|---|
| kwˈaɪə | ˈkwaɪəchoir /ˈkwaɪə/ | The stress mark goes at the front of the syllable, not next to the vowel. No dictionary prints it inside a cluster of consonants. |
| ɹ | rread /riːd/ | Broad transcription uses the plainest symbol in a sound’s range. The curled r is a narrower detail than this page is making. |
| ɾ | tbutter /ˈbʌtɚ/ | American English taps the t in butter. We checked that eSpeak taps only t and never d, so turning the tap back into t is safe: ladder keeps its d. |
| ɐ | əabout /əˈbaʊt/ | eSpeak writes the weak vowel two ways depending on where it sits. They are one sound, and it is the sound this site is named after. |
| ɜː | ɝbird /ˈbɝd/ | In American English the vowel in bird has an r in it. eSpeak leaves that r out. Stressed it is written ɝ, unstressed ɚ, the way Cambridge writes them. |
| ɪ at the end | ihappy /ˈhæpi/ | No English word ends in the short vowel of kit. The last sound of happy is closer to the vowel of fleece, and every dictionary since about 1990 says so. |
| ᵻ | ɪbushes /ˈbʊʃɪz/ | eSpeak has a reduced vowel of its own for the middle of bushes. Cambridge writes that with the vowel of kit. |
| ɫ | lfull /fʊl/ | The dark l at the end of full is the same letter as the clear one in leaf. The difference is automatic, so it does not need its own symbol. |
Words spelled one way and said two ways
A dictionary lookup cannot tell these apart, because nothing inside the word is different. Only the rest of the sentence is. We gave the converter 24 matched pairs and checked both halves of each one.
- 70.8% correct
- 24 test sentences
- readone wrong
Please read it now.
I read the book yesterday.
- recordone wrong
The record was broken.
Record it before you forget.
- presentboth right
She opened the present.
He will present the findings.
- windone wrong
The wind blew all night.
Wind the clock before bed.
- liveboth right
They live near the river.
Do not touch a live wire.
- desertboth right
Nothing grows in the desert.
Do not desert your post.
- objectboth right
She picked up the object.
They object to the plan.
- closeone wrong
Please close the door.
The station is close to us.
- leadone wrong
She will lead the group.
The pipes are made of lead.
- bassone wrong
He plays the bass guitar.
They caught a bass in the lake.
- tearboth right
A tear ran down her cheek.
Do not tear the paper.
- doveone wrong
A dove landed on the roof.
He dove into the water.
This is a small set, and it is a set we chose. It shows where the converter stands, not how often you will hit one of these words in your own writing.
How much of English changes between the two spellings
We transcribed the 20000 most common English words twice, once in each spelling, and counted the words that came out different.
8,639 of 20,000 words differ, and 3,328 of those differ only because of the r after a vowel. Take the r away and the two accents are much closer than their reputation suggests.
Checked against a dictionary nobody here wrote
A tool that grades its own homework proves nothing. So the American spelling was compared against CMUdict, a pronunciation dictionary built at Carnegie Mellon and maintained by other people entirely. On the 19290 most common words the two agree on 78.2% of them, sound for sound.
Where they disagree it is almost always a written convention rather than a different sound. Neutralise the one convention that is known to vary, the vowel before r, and agreement rises to 79.9%. Some real examples, ours first:
- was wʌz vs wɑz
- here hɪr vs hir
- yeah jɛh vs jæ
- gonna ɡənə vs ɡɑnə
- really riəli vs rɪli
- more mor vs mɔr
- even ivən vs ivɪn
- into ɪntʊ vs ɪntu
- anything ɛnɪθɪŋ vs ɛniθɪŋ
- before bɪfor vs bɪfɔr
- everything ɛvrɪθɪŋ vs ɛvriθɪŋ
- around ɚraʊnd vs ɚaʊnd
Nine tools do the transcription. The one that ranks first says nothing.
On 5 September 2026 the first page of Google for "english to ipa", "ipa transcription" and "phonetic spelling generator" was almost entirely small, single-purpose tools rather than large companies. We fetched the top four and looked for any audio at all. toPhonetics has none. The International Phonetic Alphabet site has none. PhoneticGen hands the job to whatever voice your operating system happens to ship. That gap is the reason this page exists.
Every speaker in Common Voice states their own accent, but a voice built from a recording can lose it on the way. So each voice reads a sentence loaded with words that have an r after a vowel, and we check the recording for whether that r survived. It is the one feature that splits the English-speaking world in two and it is measurable in the sound itself. The nine accents, and how each one was checked.
Updated 2026-09-05