VoiceDelivery

Reading Aloud

A passage scrolls past at a pace you set, and you read it out loud. Because the words are already chosen, all your attention goes to how you say them — and because the app knows the text, it can tell you exactly which ones you missed.

Format: Scrolling teleprompter
Pace: Target WPM you choose
Score: 0.65 accuracy + 0.35 pace
Evidence: Word-by-word diff

The method

Why it works

Reading Aloud trains delivery. With the content handled, you are free to focus entirely on crisp consonants, full vowels and clear phrasing — the mechanics of sounding good.

Nothing to invent
No topic to construct and no words to search for, so every scrap of attention goes to articulation and expression.
A moving teleprompter
The scroll sets a steady rhythm, training you to hold an even pace instead of rushing ahead or trailing off at the ends of sentences.
Read against the text
Because Eloqo knows the passage, it can compare exactly what you said to what was written — so skipped or slurred words surface precisely, not vaguely.

How a session runs

Step by step

1
Pick a passage and a target pace; the teleprompter scrolls to match.
2
Read along out loud. Narration audio is available if you want to hear the passage first.
3
Your take is aligned word-by-word against the passage, and every slip is marked on the text.

Evaluation

How it's scored

Accuracy comes from a weighted word alignment against the passage. Pace is measured against the target *you* chose, not a universal ideal — the teleprompter setting is the contract, and the score checks whether you kept it.

Deterministic · reading-accuracy-and-pace

Accuracy

65%

Every covered word scores 1.0 for a correct reading, 0.5 for a partial match, 0 for a skip or substitution. Restarts are charged separately.

Pace

35%

A Gaussian around your chosen target with sigma at 18% of it. Graded on full-duration WPM, because the target already budgets for the natural pauses at punctuation.

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.

Coverage

covered = everything up to the LAST recognisably read word

// Anchoring here is what makes stopping halfway a short read rather
// than a wall of “skipped” errors across the rest of the passage.

Accuracy

base    = (correct×1.0 + partial×0.5) / coveredWords × 100

perRun  = min(100 / coveredWords, 5)      // one restart ≈ one missed word
total   = min(runs × perRun, 20)

accuracy = clamp(0, 100, base − total)

Pace

sigma = 0.18 × targetWpm
pace  = 100 × exp(−(wpm − targetWpm)² / (2σ²))

Overall

overall = 0.65 × accuracy + 0.35 × pace

Details worth knowing

Numerals are aligned in spoken form, so “25” in the passage matches “twenty five” in your reading regardless of which side wrote it which way.
Hyphenated words are split on both sides, because recognisers emit “well-known” as two words and the alignment has to stay one-to-one.
A restart costs the equivalent of one missed word of the text you covered rather than a flat fee, so the penalty scales with how much you actually read — capped at 5 points per run and 20 in total.

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.

Three independent signs of a broken transcription withhold the score: a transcript under 30% of the passage's length, under 20% of the heard words matching anything in the passage, or more than 30% of covered words missing wholesale (humans misread words; a recogniser dropout deletes them).
Any of those is forgiven when a coherent read prefix exists — at least 5 covered words at 70% base accuracy or better — because that is a legitimate manual stop, not a failed transcription.

Reading the number

85–100Excellent

70–84Good

50–69Average

Below 50Needs practice

The coach review

What the AI adds

The review is delivery-focused, and is told not to grade vocabulary — the words are not yours:

Reading accuracy in prose: which words were skipped, misread, added or self-corrected.
Pace and steadiness, against roughly 120–160 WPM for expressive reading.
Expressiveness and intonation — whether it sounded engaging or flat.
The transcription is required to be verbatim and is explicitly forbidden from being reconstructed out of the reference text.

Keep going