NeuroPhonetics

See the sound. Feel the vibrations. Say it.

A way of learning the sounds of American English by hearing them, seeing them and feeling them — and the Android programs that carry it. One teaches the whole of it; the others each do one job, climbing from a sound to a sentence, and next to what it means and what it does.

The flagship · in development

Sentio

Hear it before you say it.

Sentio will teach you to hear, see and pronounce the vowels, syllables and words of American English — and put your attempt beside the recording.

  • Hear
  • See
  • Pronounce
  • vowels
  • syllables
  • words

How Sentio teaches

What that line means

Three things a learner can do with a sound, each shown here with real measurements from the family's own recordings and reference.

See the sound.

A sound you cannot see is hard to argue with. A spectrogram makes a vowel visible as two bright bands, and the chart of those bands is a map of the mouth. Phonema writes the same thing in letters, as IPA.

  • deed ˈdid

    01234kHzF1 286F2 2165

    tongue high and forward

  • had ˈhæd

    01234kHzF1 532F2 1637

    the mouth opens: F1 rises

  • not ˈnɑt

    01234kHzF1 677F2 1337

    the tongue goes back: F2 falls

The bright bands are the formants, and they are the vowel. F1 rises as the mouth opens (deed → had); F2 falls as the tongue moves back (had → not). Above each picture, the waveform; the dashed lines mark the vowel’s own F1 and F2, measured by the laboratory’s tracker. One speaker, rms (a man), each word cut from a sentence he read. Read speech, CMU ARCTIC.

Three words. Deed: F1 low, F2 high. Had: F1 high, F2 middle. Not: F1 high, F2 low. The numbers are on each picture.

The American vowels on the F1/F2 chartNine vowel classes measured from four American voices, each drawn as a ring around its center. Front vowels sit on the left (i of fleece highest, then ɪ, ɛ, æ lower); back vowels on the right (u of goose high, ʊ, then ʌ and ɑ low). The schwa ə has the widest, dashed ring in the middle.2500200015001000300500700900front of the mouthbacktongue highmouth openF2, Hz → the tongue forward or backF1, Hz ↓ifleeceɪkitɛdressætrapʌstrutɑlotʊfootugooseəcomma
The chart is a map of the mouth. Up is the tongue high, down the mouth open (F1); left is the tongue forward, right back (F2). Each ring is one standard deviation around where four American voices put that vowel; the dashed ring is the schwa, which has no place of its own. Centers in the hertz of an average mouth — nobody is judged in these numbers, only in their own.

The whole chart, with every vowel and how Polish, Spanish or Russian speakers hear it, is the vowel map.

Feel the vibration.

Some differences are easier to feel on your own body than to hear: voicing as a buzz under two fingers on the throat, aspiration as a puff on the back of the hand. None of our apps trains this yet — it is where the method is going — but you can try both right now.

  • missing ˈmɪs·ɪŋ

    s: noise with no pattern. Nothing at the throat (voicing 0.21).

  • losing ˈlu·zɪŋ

    z: the same hiss riding on a repeating wave — the vocal folds (voicing 0.75).

Put two fingers on your throat and say ssss, then zzzz. The z buzzes under your fingers; the s does not, and the tongue does the same thing for both. Above, forty milliseconds from the middle of each, said by slt (a woman); voicing is the laboratory’s periodicity score, 0 to 1. In running speech the buzz is not the whole difference: across the four voices the z was shorter than the s (median 94–110 against 130–160 ms, 303 z and 1,890 s before a vowel) and often only partly voiced. Voiced through its middle (a score of 0.5 or more): bdl 23 %, rms 0 %, clb 76 %, slt 91 % of their z, against at most 1 % of any s; in rms (a man) the z scored hardly more than his s. Read speech, CMU ARCTIC.

  • pike ˈpʰaɪk narrow

    85 ms

    A puff of breath before the voice starts: that is the ʰ.

  • spite ˈspaɪt

    15 ms

    After s, hardly a puff: the voice starts almost at once.

Hold a candle, or a strip of paper, in front of your lips. Pike bends the flame; spite hardly moves it. The amber stretch is the time from the p’s release to the voice — breath you can feel on the back of your hand, and the ʰ that Phonema’s narrow notation writes. Both words said by clb (a woman). Across the four voices, p, t and k at the start of a stressed word waited a median 62–87 ms for the voice (1,614 words); after s, 16–29 ms (430). Read speech, CMU ARCTIC.

Say it.

Saying it right is more than the sounds: English is timed by its stressed syllables, and the rest are allowed to shrink. Antiphon gives you a real speaker to say it back to; Sentio will judge an attempt against what the sound should be.

together tə·ˈɡɛð·ɚ

bdl100tə260ˈɡɛð100ɚrms105305155clb140270150slt106250219
əaˈkʌp‿cupəvofˈtitea
English keeps time on its stressed syllables and lets the rest shrink. Together, as each of four speakers said it (bdl and rms, men; clb and slt, women; the median of nine times each): the stressed ɡɛð is the loudest and 2.0–2.9 times as long as the weak tə before it (bar height is loudness, width is length, in milliseconds). Over 47 such words with a schwa first (312 times said), the stressed syllable was 2.0–2.5 times as long. Syllables as Phonema divides them, times from the corpus’s own phone labels. Read speech, CMU ARCTIC. Below, a phrase in Phonema’s own notation: a and of shrink to ə and əv, and ‿ marks a word running into the next.

The apps

Nothing here is published, and nothing is for sale.

Six programs are built and in testing on our own devices, none of them on Google Play yet: Redeem Time and Mimesis are the most finished; Phonema, Lectio, Antiphon and Sentio need a NeuroPhonetics service that is not public yet. Two more — Semantica and Pragma — have just been started. Every page says plainly where its program stands, and no page has a download on it. Free is everything that works inside an app; exports and what an app sends to another are Pro, coming.

Your time

Redeem Time protects the hours that make any of it possible.

  • Redeem Time

    Notice. Limit. Redeem.

    Screen time limits and honest statistics for a phone, a tablet and a TV.

    Not published yet

Your words

Find them in books and in what you hear; see how every one is said.

  • Lectio

    Read. Click. Remember.

    Reads your books and turns a chapter into the words and phrases worth learning.

    In testing

  • Mimesis

    Listen. Click. Keep.

    Listen to talks, podcasts and your own audio, and keep the sentences you hear: click, memo, export, play.

    Not published yet

  • Phonema

    See every word, as it’s really said.

    Turns English text into IPA — broad for everyone, narrow with Pro — and exports it.

    In testing

Your voice

Say it back, and hear how close you came.

  • Antiphon

    Hear it. Echo it. Speak it.

    Cuts a recording you already have into pieces short enough to say back.

    In testing

Your meaning

The ladder’s next two steps: what a sentence means, and what it does.

  • Semantica

    Find the word that fits.

    Every sense of a word, its opposites, the words it goes with, and the near-synonyms not to mix up.

    Coming soon

  • Pragma

    Catch the hint. Answer in kind.

    Requests, politeness, implied meaning, the melody that changes it — and the everyday English of the Upper Midwest.

    Coming soon

The path through the family

One ladder, from a sound to what a sentence does. Each app takes a step or two; the last two steps are being built.

  1. sounda vowel, a consonantSentioPhonema
  2. syllableonset, vowel, codaSentioPhonema
  3. wordfound, kept, saidLectioMimesisSentio
  4. sentenceheard and said backMimesisAntiphonLectio
sound → syllable → word → sentence → meaning → use. Each step stands on the one below it. The icons are the apps that work on that step; the dashed steps are Semantica and Pragma, coming soon.

And as a road through the apps — each useful on its own:

  1. 1
    Bring the words

    From your books and from what you hear.

    LectioreadMimesislisten
  2. 2
    See the sound

    Any text, written in IPA.

    Phonema
  3. 3
    Shadow it

    Say it back, after a real speaker — and write it down.

    Antiphonshadowing, dictationLectiodictation
  4. 4
    Learn the sounds

    Taught from the first vowel: hear, see, pronounce.

    Sentiothe flagship
  5. 5
    Mean it, use it

    What it means, and what it does. Coming soon.

    SemanticameaningPragmause

Redeem Timeguards the timeAnd under all of it, the hours that make it possible.

A book is coming.

The method, written down. Tell me when it’s out.

NeuroPhonetics LLC also offers electronic-structure (DFT) calculations and statistics for quality — Lab & consulting.

Private by design

No account, no ads, no analytics, no tracking, in any of the apps. Redeem Time and Mimesis never call a server at all: everything they do runs on your device, and a copy leaves it only when you choose a folder, your own Google Drive or another of your devices on your own Wi-Fi.

Phonema, Lectio, Antiphon and Sentio are built to do their work on a NeuroPhonetics service, because turning text into IPA, reading a book or cutting a recording into clips is more than a phone can do. None of those services is open to the public yet — which is the other reason those apps are not released. The rule they are built to: what an app sends is answered and not stored. Antiphon's lessons come back to your phone and the recording is not kept; your own voice is scored and deleted in the same request. Each app's privacy policy says exactly what leaves the phone and when: Redeem Time, Mimesis, Phonema, Lectio, Antiphon, Sentio.

This website is the same: no cookies, no analytics, no fonts or scripts loaded from anyone else.