What accent does this IPA transcriber use?

General North American citation forms. The CMU Pronouncing Dictionary lists American English pronunciations, and the page maps its 39 phonemes to broad IPA, so butter comes out as bʌtɚ and bird as bɝd. It has no British, Australian or other accent, and it does not show how words change in connected speech.

Why does it ask me to pick a pronunciation for some words?

The dictionary has more than one entry for many words: record has three (ɹəkɔɹd, ɹɛkɚd and ɹɪkɔɹd), and common words such as the, a and and have two or three. The page never picks one from the meaning of your sentence. Until you pick, the transcription shows the different IPA in braces; your pick then applies everywhere that word appears. Where every entry gives the same IPA, as with this, the IPA is shown straight away and only the ARPAbet and stress wait for your pick.

What happens to names and words that are not in the dictionary?

They are marked unknown: the transcription keeps the word as you typed it, in square brackets, and lists it under the IPA and in the readout. Nothing is worked out from the spelling. The snapshot holds 126,052 words, including many names, but new names, rare words, numbers written as digits and words from other languages are not in it.

Why are there no stress marks like ˈ in the IPA?

A stress mark goes before a syllable, and the dictionary gives stress per vowel without saying where each syllable starts. Rather than guess the syllable breaks, the page shows each vowel's dictionary stress in its own column: 1 for primary, 2 for secondary and 0 for unstressed, so banana is bənænə with stress ə0 æ1 ə0.

How does it decide between ə and ʌ, or ɚ and ɝ?

From the dictionary's stress digit. Unstressed AH (AH0) becomes ə and unstressed ER (ER0) becomes ɚ; AH and ER with stress 1 or 2 become ʌ and ɝ. So about is əbaʊt, but is bʌt and butter is bʌtɚ.

Are contractions and hyphenated words handled?

Yes, when the dictionary has them as written: don't, o'clock and mother-in-law are looked up whole. A hyphenated form the dictionary does not list is marked unknown rather than split into its parts; type the parts with a space between them to look each one up.

Is my text sent anywhere?

No. The 3.6 MB dictionary and the 1 kB IPA table download from this site the first time you press Transcribe or Try it, their SHA-256 fingerprints are checked, and every lookup runs in a background worker in your browser. Your draft is kept in this browser, and Start over forgets it.

How long can the text be?

Up to 10,000 characters and 2,000 words per transcription. Longer text is refused with a message rather than cut short, and it stays in the box so you can split it.

What do the exports contain?

Copy and .txt give the transcription as shown: in running text, your text with each word replaced by its IPA and every punctuation mark and line break kept; in word list mode, one line per word with a tab before its IPA. The .csv has one row per word in order, with the dictionary form, status, the entry you picked, IPA, ARPAbet, vowels and vowel stress, saved as UTF-8 with a byte-order mark so spreadsheet apps show the IPA symbols. Print prints the transcription and its word table.

IPA Transcriber

Paste English and get broad North American IPA for every word, looked up in a frozen snapshot of the CMU Pronouncing Dictionary (126,052 words) and mapped through our own 39-phoneme table. What it does differently: nothing is guessed. A word the dictionary lacks, such as a new name, is marked unknown in brackets instead of being worked out from its spelling, and a word with several entries, like record, shows every one until you pick, with one pick applied everywhere it appears. The ARPAbet and each vowel's dictionary stress sit beside the IPA, your punctuation stays in place, and the result copies, saves as .txt or .csv, or prints. The 3.6 MB dictionary downloads from this site the first time you press Transcribe or Try it; your text stays in your browser.

text0 words, 0 characters

limit: 10,000 characters, 2,000 words
source: CMUdict, 2026-09-21 snapshot, 126,052 words, BSD 2-clause
phones: 39, broad North American. AH0 ə, ER0 ɚ; otherwise AH ʌ, ER ɝ
marks: no stress marks or syllable breaks in the IPA; vowel stress 0, 1, 2 has its own column

IPAappears here

Type or paste English and press Transcribe. The IPA appears here, with every dictionary entry for words that have more than one. The 3.6 MB dictionary downloads from this site on the first press, and your text stays in this browser.

Words

Transcribed

not yet

Unknown

not yet

Variant picks

not yet

Each word is looked up as you typed it, in lower case and with curly apostrophes made straight: don't and mother-in-law are found as they are, and a hyphenated form the dictionary lacks is marked unknown rather than split into parts. An apostrophe at the edge of a word counts only when the dictionary has that form, so 'cause and goin' are found and quotation marks stay punctuation. Braces hold the different IPA of a word that is waiting for your pick. Running text keeps your punctuation and line breaks; Word list gives one line per word, the word, a tab, then its IPA. Copy and .txt give that text, .csv gives one row per word with its ARPAbet and vowel stress, and Print prints only the transcription and its table.

What comes out

cat ship thin this judge singone IPA for each word
kæt ʃɪp θɪn ðɪs dʒʌdʒ sɪŋthis has two dictionary entries, DH IH1 S and DH IH0 S; both are ðɪs, so only the stress waits for your pick.
bird bananavowel stress in its own column
bɝd bənænəbɝd has vowel stress 1; bənænə has 0, 1, 0, one digit for each vowel.
record qzxqzxqz.3 dictionary entries, then a non-word
{ɹəkɔɹd|ɹɛkɚd|ɹɪkɔɹd} [qzxqzxqz].record lists all three entries and waits for your pick; qzxqzxqz stays unknown and the full stop keeps its place.

Accuracy. Broad North American citation forms from a frozen snapshot of the CMU Pronouncing Dictionary (retrieved 2026-09-21), mapped through this page's own 39-phoneme table: unstressed AH is ə and unstressed ER is ɚ, otherwise AH is ʌ and ER is ɝ. The IPA leaves out length, syllable breaks and stress marks, and the dictionary's vowel stress (0, 1 or 2) is shown in its own column. It knows no British or other accent, no connected-speech changes and nothing about context, so a word with several dictionary entries shows every one until you pick, and a word the dictionary lacks stays marked unknown instead of being guessed. Pronunciations: the CMU Pronouncing Dictionary (cmudict), Carnegie Mellon University, snapshot of 2026-09-21, 126,052 words and 135,166 pronunciations, used under its BSD 2-clause license; source at github.com/cmusphinx/cmudict. IPA mapping: this page's own 39-phoneme table. Your text never leaves this page: the only downloads are those two files, from this site. Your draft and the mode are kept in this browser, and the Start over button at the top of the page forgets them.

Common questions

What accent does this IPA transcriber use?: General North American citation forms. The CMU Pronouncing Dictionary lists American English pronunciations, and the page maps its 39 phonemes to broad IPA, so butter comes out as bʌtɚ and bird as bɝd. It has no British, Australian or other accent, and it does not show how words change in connected speech.
Why does it ask me to pick a pronunciation for some words?: The dictionary has more than one entry for many words: record has three (ɹəkɔɹd, ɹɛkɚd and ɹɪkɔɹd), and common words such as the, a and and have two or three. The page never picks one from the meaning of your sentence. Until you pick, the transcription shows the different IPA in braces; your pick then applies everywhere that word appears. Where every entry gives the same IPA, as with this, the IPA is shown straight away and only the ARPAbet and stress wait for your pick.
What happens to names and words that are not in the dictionary?: They are marked unknown: the transcription keeps the word as you typed it, in square brackets, and lists it under the IPA and in the readout. Nothing is worked out from the spelling. The snapshot holds 126,052 words, including many names, but new names, rare words, numbers written as digits and words from other languages are not in it.
Why are there no stress marks like ˈ in the IPA?: A stress mark goes before a syllable, and the dictionary gives stress per vowel without saying where each syllable starts. Rather than guess the syllable breaks, the page shows each vowel's dictionary stress in its own column: 1 for primary, 2 for secondary and 0 for unstressed, so banana is bənænə with stress ə0 æ1 ə0.
How does it decide between ə and ʌ, or ɚ and ɝ?: From the dictionary's stress digit. Unstressed AH (AH0) becomes ə and unstressed ER (ER0) becomes ɚ; AH and ER with stress 1 or 2 become ʌ and ɝ. So about is əbaʊt, but is bʌt and butter is bʌtɚ.
Are contractions and hyphenated words handled?: Yes, when the dictionary has them as written: don't, o'clock and mother-in-law are looked up whole. A hyphenated form the dictionary does not list is marked unknown rather than split into its parts; type the parts with a space between them to look each one up.
Is my text sent anywhere?: No. The 3.6 MB dictionary and the 1 kB IPA table download from this site the first time you press Transcribe or Try it, their SHA-256 fingerprints are checked, and every lookup runs in a background worker in your browser. Your draft is kept in this browser, and Start over forgets it.
How long can the text be?: Up to 10,000 characters and 2,000 words per transcription. Longer text is refused with a message rather than cut short, and it stays in the box so you can split it.
What do the exports contain?: Copy and .txt give the transcription as shown: in running text, your text with each word replaced by its IPA and every punctuation mark and line break kept; in word list mode, one line per word with a tab before its IPA. The .csv has one row per word in order, with the dictionary form, status, the entry you picked, IPA, ARPAbet, vowels and vowel stress, saved as UTF-8 with a byte-order mark so spreadsheet apps show the IPA symbols. Print prints the transcription and its word table.

Broad North American citation forms from a frozen snapshot of the CMU Pronouncing Dictionary (retrieved 2026-09-21), mapped through this page's own 39-phoneme table: unstressed AH is ə and unstressed ER is ɚ, otherwise AH is ʌ and ER is ɝ. The IPA leaves out length, syllable breaks and stress marks, and the dictionary's vowel stress (0, 1 or 2) is shown in its own column. It knows no British or other accent, no connected-speech changes and nothing about context, so a word with several dictionary entries shows every one until you pick, and a word the dictionary lacks stays marked unknown instead of being guessed.