VoiceFluency

Monologue

A random topic, a gentle clock, and nothing to hide behind. Monologue is the closest practice here to a real conversation, meeting or talk — and it is the exercise the app's flagship scoring methodology was built for.

Length: 30 s · 60 s · 120 s
Topic: Random, with difficulty levels
Methodology: monologue-deterministic-v2
Score: Pace + cleanliness

The method

Why it works

Monologue trains spontaneous speech — forming and delivering a thought in real time, with no script. It is like going to the gym for your speaking muscles: you build the ability to talk smoothly on any topic, even the weird ones like why you would adopt a velociraptor.

A gentle time limit
Speaking against the clock stops you over-planning and makes you retrieve words as you go — and that retrieval under light pressure is what builds fluency.
Random, unfamiliar topics
With no rehearsed lines, you practise structuring thoughts on the fly. Over reps it becomes automatic, so coherence holds even when the subject is brand new.
Short, frequent reps
Low-stakes repetition calms the pressure response. Confidence grows the way any exposure-based skill does — a little discomfort, often.

How a session runs

Step by step

1
Pick a length: 30, 60 or 120 seconds.
2
Take the topic you are dealt, or draw another. A suggested story arc is available if you want one.
3
Tap record and talk until the clock runs out — or stop early when you have made your point.
4
The deterministic score lands immediately; a coach review adds the qualitative half.

Evaluation

How it's scored

Two components, both computed on-device from your transcript and the audio evidence. Given the same take, the same numbers come out every time — every parameter below is pinned to this methodology version rather than read live, so a stored score stays reproducible.

Deterministic · monologue-deterministic-v2

Pace

35%

An asymmetric Gaussian on your articulation rate. The slow side is wider than the fast side on purpose: deliberate slowness is tolerated more than rushing, while a genuinely low pace still costs you.

Cleanliness

40%

Hesitation markers per 100 words and long pauses per minute, both with an exposure floor so a short take is not extrapolated into an unstable rate.

The formula

Transcribed from the scorer that ships in the app. Every constant is pinned to this methodology version, so the same take always produces the same number.

Pace

delta = wpm − target
sigma = delta < 0 ? sigmaSlow : sigmaFast
pace  = 100 × exp(−delta² / (2σ²))

EN   target 135   sigmaSlow 45   sigmaFast 32
RU   target 115   sigmaSlow 40   sigmaFast 27

Cleanliness

effectiveWords   = max(wordCount, 50)
effectiveMinutes = max(durationSeconds / 60, 1.0)

FR = fillerCount / effectiveWords × 100     // fillers per 100 words
PR = pauseCount / effectiveMinutes          // pauses per minute

cleanliness = clamp(0, 100, 100 − 6×FR − 8×PR)

Overall

overall = (0.35 × pace + 0.40 × cleanliness) / 0.75

Details worth knowing

Only unambiguous hesitation markers reduce your score — “uh”, “um”, “er”, “erm”, “hmm” and their Russian equivalents. Ambiguous discourse words like “like”, “so”, “well”, “ну” or “типа” are counted in the statistic you see but carry no penalty: deciding whether they were fillers takes context, and that is the coach review's job.
The exposure floors matter more than they look. One pause in a 30-second take costs at most 8 points instead of 16, because the denominator is floored at one minute.
The weights are renormalised over the two components that have a deterministic local signal; there is no diction component in the local score yet.

When no score is given

A weak guess is worse than no number, so the overall is withheld rather than estimated. What was measured stays visible either way.

Fewer than 10 recognised words → no overall.
Less than 15 seconds of active speech → no overall.
When a gate trips, the raw metrics are still shown — you see what was measured, it just never becomes a number that feeds a trend.

Reading the number

85–100Excellent

70–84Good

50–69Average

Below 50Needs practice

The coach review

What the AI adds

The coach review covers everything a formula cannot:

Clarity, fluency, vocabulary and confidence, plus delivery dynamics — volume stability, pitch variation, phrase endings — judged from the audio, not from your personality.
Filler words in context, separating a genuine hesitation from ordinary usage.
Structure and logic: whether there is a real opening, a developed middle and a deliberate close, and how to reorganise your own points if there is not.
Vocabulary suggestions, but only for generic words you actually repeated — a single-use word is never called overused.

Keep going