Praat for English Learners: See Vowels, Pitch, and Timing Without Becoming a Phonetician

Use Praat for one pronunciation target at a time: compare a vowel, pitch contour, or timing pattern without treating acoustic graphs as a score.

Maya Bennett

FunFluen editor · Everyday conversation

Aug 29, 2026 8 min

The short answer

Use Praat as a magnifying glass, not a judge: choose one audible question, show only the analysis that can help answer it, compare cautiously, then return to speaking.

Open Praat for the first time and one vowel can look like a tiny weather system: waveform, red dots, blue line, grey bands, numbers in Hz. Good news: you do not need to understand the whole climate. You need one question and one display.

Pick the target before you look at the graph

I am checking one vowel

Use the waveform/spectrogram to locate the vowel. If you have a clear reason to compare vowel quality, show the formant tracks and look at the broad pattern in the same vowel and similar context. Do not hunt for a universal F1/F2 score.

I am checking one pitch movement

Show the pitch contour. Compare shape: where does the voice rise, fall, peak, or flatten? Do not try to make your absolute Hz numbers equal another speaker’s.

I am checking timing

Select the same meaningful unit in both recordings—a vowel, stressed syllable, word, or short phrase—and compare the broad timing difference. Do not turn a useful “mine is much longer here” observation into a mission to match someone’s milliseconds.

If you cannot say what audible feature you are checking or what speaking change you might make, do not measure it yet.

The 90-second Praat setup

As of August 29, 2026, Praat’s official site recommends version 7.0.02 and lists spectrogram, pitch and formant analysis among its current speech-analysis tools. The official recording tutorial gives a simple path:

In the Objects window, choose New → Record mono Sound....
Record one short model or your own attempt, then Stop and Play it.
Choose Save to list.
Select the Sound object and choose View & Edit.

You can also open an existing sound file you are allowed to use. Keep the sample short. One word or one short phrase is enough for this workflow.

Praat’s SoundEditor can show spectrogram, pitch and formant analyses. Resist the urge to switch everything on just because the menu exists. Your laptop has many settings too; you do not adjust all of them before sending an email.

Lane 1: use formants as a vowel clue, not a magic coordinate

A formant is part of the spectral structure Praat estimates across time; Praat’s Formant object stores contour information including formant frequencies and bandwidths. For learners, the first two formant patterns can sometimes provide a useful clue about vowel quality.

That does not mean there is one perfect pair of numbers that every English learner should copy. A peer-reviewed review of English vowel measurements reports substantial variability across speaker groups, speech samples and measurement points. Praat itself goes further: its formant-analysis FAQ says different settings can produce different results and that the LPC-based analysis comes without a guarantee.

A beginner-safe vowel comparison

Use the same word or carrier phrase for model and learner attempt.
Record them in reasonably similar conditions if possible.
Zoom into the vowel in View & Edit.
Listen first. What difference are you trying to understand?
Show the formants and compare the broad relationship or movement—not just one isolated red dot.
Make one articulatory change, rerecord once, and listen again.

Useful question: “Does my second attempt move toward the model in the way I expected, and does it also sound closer?” Less useful question: “What exact Hz number makes this vowel correct?”

Lane 2: compare pitch contour shape, not another person’s voice

Praat can display a pitch contour—a changing estimate of periodicity across time. For ordinary intonation work, Praat’s current pitch-analysis guidance recommends filtered autocorrelation. The important beginner warning is in the pitch-settings documentation: pitch floor and top affect the analysis, and speakers can vary enormously in pitch range.

So if your model’s question rises from a lower number to a higher number, your goal is usually the movement and placement, not the same absolute values.

Try the contour trace

Record the same short sentence twice: once as a statement and once with the intonation target you are practising. Show the pitch contour and answer only these questions:

Where does the main rise or fall begin?
Which word contains the biggest pitch movement?
Does the contour keep moving where yours becomes flat?

If Praat drops the line through a clearly voiced part, do not immediately “fix” your pronunciation. Praat’s pitch documentation explains that analysis settings can affect what is tracked. Listen, zoom, and recheck first.

Lane 3: compare timing only at a meaningful boundary

Praat’s time-domain tools display a sound’s start, end and total duration, and the editor lets you select a stretch of audio. That makes timing comparisons wonderfully simple—as long as you compare the same thing.

Suppose your target is a stressed vowel in a phrase. Select that vowel in the model, then the equivalent vowel in your attempt. Or if your problem is a long pause, select the pause in both recordings. You are looking for a useful difference such as:

“I am holding this stressed vowel much longer.”
“My pause comes before the thought group instead of after it.”
“I rush the final two words compared with the rest of my phrase.”

You are not trying to turn the model’s duration into a universal law. Different speakers, rates and meanings naturally produce different timings.

Three times the graph should lose the argument

The graph looks dramatic, but you cannot hear the difference

Do not create a practice target just to satisfy the picture. Ask a listener, check another recording, or choose an audible feature. The graph is evidence, not a grade.

You and the model are different speakers with very different pitch ranges

Compare contour direction and where the movement happens. Do not match absolute F0 values across speakers as if you were tuning two identical instruments.

Your recording conditions or analysis settings are inconsistent

Re-record or standardize the comparison before drawing a pronunciation conclusion. Praat’s formant and pitch analyses are estimates affected by settings; noisy evidence deserves less confidence, not more decimal places.

The two-attempt repair loop

Praat is most useful when the analysis ends quickly.

Name one audible target in plain English: “My rise starts too early,” “This vowel still sounds too close to my other vowel,” or “I hold this word much longer.”
Record attempt A.
Use one relevant display to check the hypothesis.
Make one speaking change.
Record attempt B.
Listen to A, B and the model. Use the graph only to answer the original question.
Close or minimize Praat and produce the target again without watching it.

If B sounds better and the visual clue moves in the expected direction, you have a useful practice hypothesis. If the graph moves but the speech does not sound more appropriate or understandable, reconsider the target.

Talk about pitch naturally, too

Original: “I need to do my pitch higher here.”

Classification: unusual/overly formal/non-idiomatic.

Listener interpretation: understandable, but do my pitch higher is not the natural collocation.

Learner intent: increase pitch or create a larger rise at one point.

Natural alternatives: “I need to raise my pitch here,” “I need a bigger rise here,” or “My pitch should rise later.”

Context note: in technical phonetics, you can say “the F0 is higher,” but ordinary pronunciation coaching usually uses raise/lower your pitch, a rising/falling contour, lengthen/shorten the vowel, and hold the word longer.

After the graph finds a real target, practise the sound—not the software

If your Praat check identifies a concrete feature in a line from a supported video, FunFluen can help you repeat that line, adjust playback speed, navigate sentence by sentence, and move into speaking practice where the required audio/subtitle context is available. That can reduce practice friction after diagnosis.

FunFluen does not run Praat analysis, measure your formants or F0, validate acoustic measurements, or perfectly score pronunciation.

Review FunFluen for contextual pronunciation practice. The first step after the click is to review the extension listing before installing.

The successful Praat session ends with the graph closed

Before you finish, say this aloud:

“I am changing ___, not trying to match ___.”

For example: “I am moving the pitch peak later, not trying to match her Hz.” Or: “I am shortening this pause, not trying to copy his exact milliseconds.”

Then say the phrase again without looking at Praat. If you can hear and produce the target more reliably, the graph did its job.

Two Praat questions learners actually need

Do I need to match the model’s exact Hz or formant numbers?

No. Speakers differ, formant measurements show substantial variability, and Praat’s analyses depend on settings. Use exact values as measurements inside a controlled comparison when they answer a specific question—not as universal pronunciation targets.

What if Praat’s red dots or pitch line look obviously wrong?

Trust neither the display nor your anxiety automatically. Listen, zoom in, recheck the relevant settings, and rerecord if necessary. If the tracking still looks implausible, do not build a pronunciation repair around it.

Sources

Explore more language-learning guides in Media-Based Language Learning.