Your voice has controls: the speaking skills worth learning on purpose

Ptichi's working rule is deliberately small:

Change one control. Compare. Keep the change only if the listener-relevant result gets better without extra strain or artificiality. Then try new words.

That last part matters. A beautiful second take of the same memorized sentence is practice. It is not yet a skill you can use in an interview, meeting or story.

You do not need Ptichi to try the exercises below. An ordinary recorder is enough. The desktop app is still being developed, so this page does not pretend that every control already has automatic analysis.

1. Breath for speech: enough air, not maximum air

"Speak from the diaphragm" sounds concrete until you try to turn it into an instruction. The diaphragm is involved in breathing, but a microphone cannot tell you that you used it "correctly", measure lung volume or prove that a deeper inhale made the sentence better.

A more useful target is respiratory-speech coordination: replenish before a thought when you need to, take enough air for that thought, and keep the useful ending available. Voice production depends on coordinated respiration, phonation and resonance rather than one magic body part. ASHA's overview of voice production and voice disorders is a useful reference for that broader model.

Try this

Take one sentence with a clear thought:

The migration finished yesterday, and the reconciliation starts this morning.

Record it normally. Then record it again after one quiet, comfortable replenishment before the thought. Do not inhale to maximum. Do not hold the breath to make the preparation feel important.

Listen for two things: did you finish the phrase comfortably, and did breathing become less noticeable rather than more noticeable?

If the second take creates dizziness, pressure, throat strain or a ritual of giant breaths, it is the wrong direction. Bigger breathing is not a progress metric.

Read the full guide: Breath for speech — why “take a deeper breath” is usually the wrong instruction.

2. Thought groups: let the listener hear the structure

A pause can help, but "pause more" is weak advice. The real job is to make the relationship between ideas recoverable.

Consider:

We keep the old report for reconciliation. New decisions use the updated report.

The listener needs to hear two purposes. A useful boundary may involve silence, final lengthening, a pitch movement or reset, or a combination of cues. Research on prosodic boundaries supports treating the boundary as more than a silent interval alone. One controlled study of phrase-boundary cues manipulated pause, final lengthening and pitch information.

This does not mean "make every boundary longer". A pause cannot fix missing context, jargon or a sentence whose logic is broken.

Try this

Record a dense two-part sentence as one stream. Then record it again while deliberately finishing the first idea before beginning the second. Do not choose a duration target.

Listen without the script and ask: where does idea one end? If practical, ask another person the same question.

Then use a new sentence. If the technique only works where you memorized the pause, you learned a location, not the skill.

Read the full guide: Thought groups — why a useful pause is about meaning, not milliseconds.

3. Prominence: decide what the listener should notice

The same words can point to different meanings depending on which item becomes prominent.

WE need the new version.

We need the NEW version.

The first can correct who. The second can correct which version.

Prominence is relative. If every important word gets maximum pitch, loudness and duration, nothing is especially prominent anymore. The goal is not to create a larger graph. The goal is to make the intended focus easier to recover.

Try this

Use the sentence:

We need the final file today.

Say it three times, each time correcting a different imagined misunderstanding: we, final, today.

Listen back without looking at your notes. Can you tell which correction you intended? If the target word sounds shouted or theatrical, reduce the magnitude. Keep the contrast, lose the performance.

Read the full guide: Prominence — make the important word land without sounding theatrical.

4. Finality: let a finished statement sound finished

This is one of the most useful controls for English, especially if your habitual endings drift upward.

A definite English statement commonly uses a flat or falling final contour. A yes/no question commonly rises. Wh-questions commonly fall. A fall-rise can carry reservation, politeness, checking or a sense that more is coming. Cambridge's overview of English intonation gives practical examples of these patterns.

The important word is commonly. "Always go down at the end" is not a good rule. Intonation carries pragmatic meaning, and dialects differ. The technique is to choose finality when you actually mean finality.

Also separate two things that often get confused:

falling pitch is a contour;

fading away is losing audibility or clarity.

A good final fall does not require the final useful word to disappear.

Try this

Record the same sentence three ways:

You sent the final file.

1. A definite confirmation: you know it happened.
2. A genuine question: you do not know whether it happened.
3. A checking/surprised version: you think it happened, but you are verifying it.

Do not use arrows while recording. Decide the speech act first, then listen to what the contour actually did.

For a definite statement, practise letting the ending settle without forcing your pitch to the bottom of your range. If the voice becomes unnaturally low, creaky or quiet, you overshot.

Read the full guide: Statement finality — stop a finished thought sounding accidentally unfinished.

5. Question versus statement: intonation should carry the speech act

It is tempting to learn intonation as a diagram: statement arrow down, question arrow up. That can be useful for noticing a contrast, but it is not enough for real speech.

A speaker can ask a question with a falling contour, use a rise while checking shared information, or use a fall-rise to signal that the sentence is not the whole story. The useful skill is not drawing the right arrow. It is making the communicative action recoverable.

Try this

Use another neutral line:

The review starts Monday.

Produce it as:

• a firm statement;
• a genuine yes/no-style check;
• a surprised confirmation;
• a sentence that clearly signals "and there is more".

After a short break, listen in random order and label what each version sounds like. If you cannot recover your own intent, exaggerate the contrast once, then reduce it until it sounds natural.

For Russian and German, this exercise needs language-specific adaptation. Ptichi should not silently inherit an English intonation lesson into another language.

Read the full guide: Question vs statement intonation — learn intent, not arrows.

6. Pace: control information locally, not with one ideal WPM

Words per minute can describe a recording. It cannot, by itself, tell you whether the listener recovered the important point.

A familiar story and a dense status update can have the same average rate and create very different listening demands. Slowing everything may help one problem and create another: the message can become flat, fragmented or simply longer.

A better control is local listener time. Keep a natural baseline, then give a little more time where information becomes dense, unfamiliar or consequential.

Try this

Use a 20-second update with one critical item — a date, decision, risk or next step.

Record:

1. your normal version;
2. a globally slower version;
3. a version at roughly your normal pace with one deliberate boundary or local slowdown around the critical item.

Do not choose by WPM alone. Ask which version makes the critical information easiest to recover without sounding labored.

This is why Ptichi treats pace measurements as descriptive evidence unless a listener/task consequence justifies a stronger claim.

Read the full guide: 145 words per minute is not a communication strategy.

7. Phrase endings: do not disappear before the useful word lands

A speaker can make the right intonation movement and still lose the ending. Common failure modes are different:

• pitch falls appropriately, but energy collapses;
• the final words accelerate;
• articulation becomes vague;
• the phrase was simply too long for the speaker's current breath organization.

The correction is not "get louder at the end". The job is to preserve the final critical item.

Try this

Say:

The next step is Monday.

Then:

The next step is a security review.

Record one ordinary take and one take where you deliberately keep the final critical item available. Listen without the script. Can you recover the last useful words easily? Does the second version stay comfortable?

If the ending becomes punched or over-loud, reduce the intervention. Stability is not emphasis at any cost.

Read the full guide: Do not disappear before the important word lands.

8. Clear articulation: recover the word, do not erase the accent

Clear speech is not the same as native-like speech.

If a listener repeatedly misses the critical name, number or technical term, articulation can be a useful control. The bounded goal is simple: make the item recoverable with enough consonant/vowel contrast for the task.

Accent and dialect are not errors. Ptichi should not turn accent similarity into a quality target, and "move your mouth more" is too vague to be a universal fix.

Try this

Choose a phrase with one critical item:

We need the rollback plan before noon.

Record it. Hide the script and replay it later, or ask another listener to repeat only the critical item. Then make one careful clarity change and repeat the test.

After that, change the wording and the critical item. If your improvement disappears immediately, the first success may have been memorized over-articulation rather than transferable control.

Read the full guide: Clear articulation without accent removal.

9. Resonance and ease: "less nasal" is not a universal instruction

"Stop sounding nasal" is another piece of advice that sounds more precise than it is.

Nasal resonance is normal for nasal sounds. Oral/nasal balance changes with the intended sound and varies across languages and dialects. ASHA's resonance-disorders overview also makes an important distinction between ordinary resonance choices and conditions such as hypernasality or hyponasality that can involve structural or physiological factors.

That means Ptichi should not listen to an ordinary consumer recording and diagnose a resonance disorder. Nor should it teach "less nasal" as a global quality score.

A safer non-clinical direction is ease plus clarity: compare an effortful/squeezed production with an easier one, and keep the change only if speech remains clear and natural.

Try this

Record a short comfortable phrase. Then make one version that feels deliberately squeezed, only briefly enough to recognize the contrast, and one version where you release unnecessary effort while keeping the words clear. Do not chase a particular vibration sensation.

Ask: did the easier version remain clear, and did effort actually drop?

Humming, straw phonation and related semi-occluded exercises can be useful in voice work, but they are not a universal foundation or an intensity contest. Setup and dose matter. Persistent pain, worsening hoarseness, strain or unusual resonance concerns are not a reason to train harder.

Read the full guide: “Sound less nasal” is not a complete voice instruction.

10. Projection: audible without pushing

Before coaching projection, separate quiet speech from a quiet recording.

If the microphone gain, distance, input choice or processing is wrong, telling the speaker to push harder treats the person as the fix for a capture problem. This is exactly why Ptichi separates capture truth from speaking behavior.

When delivery really is too quiet for the context, the goal is easy audibility. Loudness is not confidence. Lower pitch is not authority. Throat force is not presence.

Try this

First make a normal recording at a sensible, repeatable microphone setup. If the input is clearly too low, fix the setup before changing your body.

Then record the same short explanation twice: ordinary delivery and one comfortably more audible version. Compare audibility, clarity, naturalness and effort. Do not choose the take simply because its waveform is bigger.

If throat pressure or discomfort rises, stop increasing output.

Read the full guide: Voice projection starts with one awkward question: is the microphone the problem?.

What these controls have in common

Most bad voice advice starts with a useful observation and turns it into a monotonic target:

breathe deeper; pause more; lower the ending; vary pitch more; slow down; resonate more; get louder.

That is where a technique becomes a caricature.

Our read

The more useful training model is almost the opposite: use the smallest change that creates the intended listener effect, then check whether it transfers without added effort or artificiality.

This is why Ptichi should keep different kinds of evidence separate.

A self-listening task can tell you whether you hear the contrast. An acoustic measure can describe what changed. A listener task can test whether the important information became easier to recover. None of those automatically replaces the others.

And sometimes the right result is: Take B was worse. That is useful information. You found an overcorrection before carrying it into a real conversation.

Transfer is the test

Use this loop for every technique on this page:

Take A → one change → Take B → compare → new words.

If the change only survives on the sentence you practised five times, keep calling it practice. When you can reproduce the useful behavior on a new question, explanation or story — with less prompting and without extra strain — it starts becoming a skill.

Ptichi is being built around that distinction. The website can already give you bounded exercises; the desktop recording and checking path is developed separately and must earn each automatic claim rather than pretending every visible graph is coaching.

For now, choose one technique above or continue with one thought-boundary contrast. Both work with an ordinary recorder and do not require a global voice score.