English pronunciation, made visible.
SayNative turns any English text into a visual pronunciation guide — every word annotated sound by sound, with stress, syllables, silent letters and the rhythm of natural connected speech.
Who it's for
Learners who can read English but aren't sure how it should sound; teachers preparing reading material; parents helping children decode tricky words; and anyone curious why knight and night sound the same. Paste your text, and read along with the marks.
How it works
SayNative is powered by a grapheme-to-phoneme (G2P) engine that predicts pronunciation algorithmically rather than looking words up in a fixed list. The models are trained on the CMU Pronouncing Dictionary: a joint n-gram model handles open-vocabulary words, while decision trees handle function words and align each sound back to the letters that spell it. That alignment is what lets the tool mark silent letters, underline digraphs like th, and place stress on the right part of the word.
Because the output is predicted, unfamiliar names and rare words get a best-effort guess — those are flagged with a small ? so you know to double-check. Words with more than one pronunciation (like read) are marked with a ².
The accent
Transcriptions follow General American English, including natural connected-speech features: flapped t (as in water), weak forms of function words, and linking between words.
Who makes SayNative
SayNative is an independent, one-person project — there is no company behind it, and mail sent to the site goes straight to the engineer who builds it. It grew out of a personal G2P (grapheme-to-phoneme) research project and a simple frustration: pronunciation is the one part of English a learner can't check alone from text. The site is free, has no accounts, and is funded by reader donations.
How the content is made
The lesson series follows what pronunciation research actually supports: a systematic sound-by-sound sequence (the approach validated by the National Reading Panel's phonics findings), and listening practice with varied voices — High-Variability Phonetic Training, one of the most replicated results in second-language pronunciation research (Thomson 2018; Sakai & Moorman 2018). Every example word's notation sheet is generated by the same engine that powers the tool, so the lessons and the tool never disagree with each other.
Articles are written for this site — not syndicated, not generated en masse — and every pronunciation claim in them can be checked against the tool itself: paste the example words and read the marks.
Corrections
Because pronunciations are predicted by a model, mistakes are possible — and reader reports are how they get fixed. If you find a wrong pronunciation or an error in a lesson or article, tell us: reports are checked against the CMU reference dictionary and native-speaker recordings, and fixes ship to the engine or the content, usually within days.
Get in touch
Found a wrong pronunciation, or have an idea? We'd love to hear it — see the contact page or write straight to hello@saynative.com. If SayNative helps you, you can also support the project.