Voice Spectrogram Activity for Kids: See Your Voice

Whistle at a spectrogram and you get one thin line. Sing the same note instead and you get a ladder: eight or ten evenly spaced lines stacked above each other. Same pitch, completely different picture, and the difference is why a whistle and a voice do not sound alike. That is the first two minutes of this voice spectrogram activity kids can run on any phone, and by the end of twenty minutes they will be able to tell “ee” from “ah” by looking rather than listening, which is roughly what Siri does.

Key Takeaways

• You’ll install a free spectrogram app and read your own voice, a whistle, a hiss and a clap off a moving picture of sound.
• It teaches that pitch is the spacing of the harmonic stack while vowel identity is which bands are bright, and that these are independent.
• Best for ages 9–15. Younger children enjoy the display; the formant comparison needs some patience.
• The usual failure: using an app that only shows a bouncing bar chart. You need a scrolling time-frequency view.
• This is the exact representation that speech recognition starts from, which makes the whole thing less abstract than it sounds.

What You’ll Need

Item	How much	Where to get it	If you don’t have it
Phone or tablet	1	Any recent one	A laptop with Audacity works and gives a bigger picture
Free spectrogram app	1	Search your app store for “spectrogram” or “spectrum analyser”; Spectroid on Android is a common choice	Audacity on a computer: record, then switch the track to Spectrogram view
Audacity or Praat, optional	1	audacityteam.org, or Praat from the University of Amsterdam	Praat is free and is what phoneticians actually use
A quiet room	1	Anywhere	Background noise fills the picture with fog
Printed list of sounds to try	1 sheet	Write it out	Whistle, aah, eee, ooo, sss, shh, clap, hum, kettle, keys
Something that makes a steady tone	1	A tuner app or a recorder	For comparing a pure tone against a voice
Pencil and paper	1 each	Anywhere	Sketch what you see; it forces real looking

Time: setup 10 min, experiment 25 min, discussion 15 min Cost: free Mess level: none, though it does get loud and silly

Do the Experiment: A Voice Spectrogram Activity Kids Can Read

Step 1 — Get a real spectrogram on screen

Install the app and set the frequency range to roughly 0 to 5,000 hertz, which is where voices live. You want a picture that scrolls sideways with time, with frequency up the side and loudness shown as colour or brightness. A single bouncing bar graph is not a spectrogram and will not work for this.

Step 2 — Whistle, and look at one line

Whistle a steady note. You should see a single bright horizontal line, and nothing else. That is what a pure tone looks like: energy at one frequency only. Slide the whistle up and down and watch the line move.

Step 3 — Sing “aah” on the same note

Now sing a vowel at the same pitch. The single line becomes a stack of lines, evenly spaced, usually eight or more. Count them. The bottom one is the fundamental, and the spacing between every line equals that fundamental.

Step 4 — Sing up a scale and watch the whole ladder move

Sing a rising scale on “aah”. The entire stack slides upward together, keeping its spacing proportional. Pitch is the spacing, and the spacing is the picture of pitch.

Step 5 — Hiss, shush and clap

Say a long “sss”. No lines at all, just a bright fuzzy band high up. Now “shh”: the same fuzz, but noticeably lower. Then clap once: a single vertical streak from bottom to top, there and gone. Three completely different shapes, three completely different kinds of sound.

Step 6 — The formant test, which is the real point

Sing “eee”, then “ah”, then “ooo”, all on exactly the same pitch. The line spacing does not change, because the pitch did not change. What changes is which lines are bright. For “eee” there is energy very low and again high up around 2,000 hertz or more; for “ah” the two bright regions sit much closer together in the middle. Those bright regions are formants, and they are what makes a vowel a vowel.

Step 7 — Go hunting round the house

Point it at a kettle, a tap, a fridge, keys jangling, a phone ringtone, a dog, a sibling. Sketch each one. Then try to guess a sound from its picture with your eyes on the screen and your ears covered.

What you should see: The whistle gives one line, sometimes with a faint second one if the whistle is not pure. A sung vowel gives a clear stack, and a child’s higher voice gives more widely spaced lines than an adult’s. The “sss” fuzz sits mostly above 4,000 hertz and the “shh” fuzz lower, around 2,000 to 3,000. The clap is a vertical line because it contains all frequencies at one instant. The formant test is the one that takes patience: the bright bands are obvious once you know to look for them, and invisible until someone points them out.

What’s Actually Going On

A spectrogram is three quantities on one picture: time along one axis, frequency along the other, and loudness as brightness. It is made by chopping the sound into short slices, a few tens of milliseconds each, and running a Fourier transform on every slice to work out how much energy sits at each frequency. Stack the slices side by side and you get a picture of how a sound’s recipe changes through time.

Your voice has two separate parts, and the picture separates them cleanly. The vocal folds chop the airflow into a buzz at some rate, say 220 times a second. A repeating buzz is not a pure tone: mathematically it contains energy at that rate and at every whole multiple of it, which is why you see a ladder rather than a line. The spacing between rungs is the rate the folds are vibrating, and that is what you perceive as pitch.

Then the sound passes through your throat, mouth and nose, which form a tube with bends and openings you can change. Like any tube, it has resonances, and it amplifies harmonics that land near them while damping the rest. Those boosted regions are formants. Moving your tongue and lips moves the resonances, which changes which harmonics get amplified, which changes the vowel. The canonical measurements are over seventy years old: Gordon Peterson and Harold Barney published the frequencies of the first two formants for ten American English vowels in the Journal of the Acoustical Society of America in 1952, and for adult male speakers their averages put “ee” near 270 and 2,290 hertz, “ah” near 730 and 1,090, and “oo” near 300 and 870. Children’s formants sit higher because their vocal tracts are shorter.

That is why the formant test in Step 6 is the important one. Pitch and vowel are carried by two different features of the same picture, and they can be varied independently. You can sing any vowel at any pitch. This is also the reason a whisper still has recognisable vowels despite having no vocal-fold buzz at all: the formants are still there, shaping a noise source instead of a buzz.

The fricatives make the opposite point. “Sss” is not a buzz at all; it is turbulent air hissing past a narrow gap between tongue and teeth, which produces broadband noise with no periodicity, so there is no ladder to see. Move the constriction back for “shh” and the resonant cavity in front of it gets bigger, so the noise shifts down. Same mechanism as the vowels, applied to noise.

One honest limitation. Every spectrogram involves a trade-off set by the slice length, and you cannot escape it. Short slices locate events precisely in time but blur frequency; long slices resolve frequency finely but smear the timing. If the app lets you change the window size or FFT length, change it and watch the picture go from sharp vertical detail to sharp horizontal detail. That trade-off is not a software limitation, it is a mathematical one, and seeing it is worth more than any single picture.

How to Explain It to Your Kid

Ages 5–8: a picture of sound

Let them make noises and watch. Tell them the picture shows low sounds at the bottom and high sounds at the top, and brighter means louder. Then play a game: you make a sound with your eyes shut and they guess it from the picture, and then swap. No vocabulary needed, and they will start noticing patterns on their own.

Ages 9–12: pitch is the spacing, voice is the pattern

Introduce the ladder and have them count the rungs. Then do the test that makes the point: sing the same note on three different vowels and see the spacing stay put while the bright bits move. The sentence to land is that pitch and vowel are two different things in the same sound, and the picture shows both at once.

Ages 13+: Fourier slices, formants and speech recognition

Give them the real framework: a short-time Fourier transform, harmonics at integer multiples of the fundamental, vocal tract resonances as formants, and the time-frequency trade-off. Then connect it forwards. Automatic speech recognition begins by converting audio into almost exactly this representation before any machine learning happens at all, which means the picture on their screen is the input format of every voice assistant they have ever used.

The question to ask: “If I whisper a vowel with no buzzing from my vocal folds at all, how could you still tell which vowel it was from the picture?”

Why Teaching This Early Pays Off

This activity makes an abstraction visible, which is rare. Frequency content is normally a thing you are told about; here it is a thing on a screen that responds to your own mouth in real time. That turns the word “frequency” from vocabulary into an object, and the same picture turns up later in physics, music technology, speech therapy, seismology, radio astronomy and machine learning.

The specific intellectual move is decomposition: taking something that feels like one thing, a voice, and seeing that it is made of separable parts that can be changed independently. Children who can do that have a much easier time with any system that has multiple interacting causes, which is most of them.

There is also a useful demystifying effect. Voice assistants feel like magic until you see the picture they actually work from. Showing a teenager that speech recognition starts with a spectrogram, not with understanding, does more for their intuition about what machine learning is than a long explanation would. Chi and Wylie’s ICAP framework would predict the gain comes from them explaining what they see rather than watching you point, so hand over the phone. As ever, one afternoon with a spectrogram does not change grades, and nobody should pretend it does. It gives a child a picture they will recognise for the rest of their life.

Before You Start: What Parents Should Know

Best age	9–15. Five-year-olds enjoy it as a toy, which is fine, but the formant idea needs about nine
Supervision	Light. The only adult job is picking a decent app and resisting the urge to drive
Real hazards	None. This is a listening and looking activity with no equipment risk. If anyone starts shouting into the microphone to get a brighter picture, note that brightness comes from the colour scale and not from volume, and turn the gain down instead
Don’t	Don’t settle for an app that shows a single bar graph of levels. Without the scrolling time axis, none of the interesting results are visible

When It Doesn’t Work

• The screen is a wall of fog. Background noise, or the app’s gain is too high. Move somewhere quieter and turn the sensitivity down until the quiet moments are dark.
• Your sung vowel shows no ladder. The frequency range is set too wide. Set the top of the scale to about 5,000 hertz instead of 20,000, and the harmonic spacing becomes visible.
• The lines are blurry smears, not sharp. The time slice is too short. If the app offers an FFT size, increase it. This is the trade-off in action rather than a fault.
• You cannot see the formants at all. Try it with a lower voice. An adult male voice has closely spaced harmonics, which outlines the formant bands much more clearly than a child’s widely spaced ones.
• The app shows frequency on a strange curved scale. Many default to logarithmic, which is good for music and bad for seeing even harmonic spacing. Switch to a linear frequency axis if you can, and the ladder becomes properly even.

Take It Further

1. Change one variable: the vowel, at fixed pitch. Record five vowels sung on exactly the same note, then measure the first two formant frequencies for each from the picture and plot one against the other. You will have reproduced the shape of Peterson and Barney’s 1952 vowel chart on a kitchen table.
2. Compare voices across the family. Everyone sings the same note on “ah”. Compare harmonic spacing, which should be nearly identical, with formant positions, which will not be. Shorter vocal tracts put formants higher, so you can rank people by vocal tract length without measuring anyone’s throat.
3. Use Praat properly. Praat, the free phonetics software from the University of Amsterdam, will overlay formant tracks and pitch contours on a spectrogram automatically. For a teenager interested in linguistics or audio, it is the real tool rather than a toy version of one.

The ladder you are looking at is the harmonic series, which is also what makes a straw instrument sound buzzy — see the straw oboe experiment. A spectrogram is also the best way to see the Doppler wobble in the swung buzzer experiment, and it makes the two tones in the sound beats experiment visible as two separate lines.

What to Watch For Over the Next 3 Months

• Week 4: They describe a sound in terms of high and low content rather than just loud and quiet.
• Month 2: The word harmonic or frequency turns up correctly about music, a voice or a machine noise.
• Month 3: They ask how a voice assistant actually recognises words, and are not satisfied with “it just does”. That question is the best thing this activity can produce, and the answer starts with the picture they already know.

Frequently Asked Questions

Which spectrogram app should we use?

Any that shows a scrolling picture with frequency on one axis and time on the other. Spectroid is a widely used free option on Android, and several free iOS apps do the same job. On a computer, Audacity works well, and Praat is what researchers use. Avoid anything that only shows a row of bouncing bars.

Why does a whistle look so different from a sung note at the same pitch?

Because a whistle is close to a pure tone, with energy at essentially one frequency, while the vocal folds produce a repeating buzz that contains the fundamental plus every whole multiple of it. Same pitch, very different recipe, which is exactly what timbre means.

What are the bright bands in a vowel?

Formants: resonances of the throat and mouth cavity that amplify whichever harmonics fall near them. Moving the tongue and lips moves the resonances, which changes which harmonics are boosted, which changes the vowel you hear.

Can we see ultrasound or very low sounds?

Only what the microphone and the app can capture, which usually tops out somewhere between 15 and 20 kilohertz and rolls off badly at the bottom. A phone microphone is optimised for voices, so the extremes are missing rather than absent from the world.

Is this the same thing speech recognition uses?

It is the starting point. Voice systems convert audio into a spectrogram-like representation, often a variant called a mel spectrogram, before any pattern recognition happens. So yes, your child is looking at the input format of every voice assistant in the house.

---

About the author

Ricky Flores is the founder of HiWave Makers and an electrical engineer with 15+ years of experience building consumer technology at Apple, Samsung, and Texas Instruments. He writes about how kids learn to build, think, and create in a tech-saturated world. Read more at hiwavemakers.com.

---

Sources

1. Peterson, G. E., & Barney, H. L. (1952). “Control Methods Used in a Study of the Vowels.” Journal of the Acoustical Society of America, 24(2), 175–184. https://doi.org/10.1121/1.1906875
2. Boersma, P., & Weenink, D. “Praat: doing phonetics by computer.” University of Amsterdam. https://www.fon.hum.uva.nl/praat/
3. Audacity Team. “Audacity: free, open source audio editor.” https://www.audacityteam.org/
4. Acoustical Society of America. “Acoustics.org: public resources on sound and vibration.” https://acoustics.org/
5. National Institute on Deafness and Other Communication Disorders. “Noise-Induced Hearing Loss.” National Institutes of Health. https://www.nidcd.nih.gov/health/noise-induced-hearing-loss
6. Chi, M. T. H., & Wylie, R. (2014). “The ICAP Framework: Linking Cognitive Engagement to Active Learning Outcomes.” Educational Psychologist, 49(4), 219–243. https://doi.org/10.1080/00461520.2014.965823
7. National Research Council. (2012). A Framework for K-12 Science Education: Practices, Crosscutting Concepts, and Core Ideas. National Academies Press. https://doi.org/10.17226/13165