PRE-LAUNCH BETA

Youdon'tlearnalanguage.Youtrainone.

If you've been grinding lessons for months and still can't hold a real conversation — you weren't failing. Your app was optimising for something else entirely.

No streaks. No fake levels. Four honest numbers — and a system intelligent enough to act on them.

What's your real fluency level? →

2 minutes · Free · Get your personalized VOICECAST™ score

74
Pronunciation
68
Comprehension
52
Production
81
Retention

You opened the app for 63 days straight.
You still couldn't order without pointing at the menu.

You know how to conjugate "être" in four tenses.
You don't know how to say "sorry, can you slow down?"

You passed every lesson.
You failed every real conversation.

The app was never trying to make you fluent.

It was trying to keep you opening it.

THE PROBLEM

You were never the problem.
Your tools were designed for something else entirely.

Duolingo has 500 million downloads. Its own research team found near-zero speaking improvement. That's not a side effect — it's the architecture. Engagement apps need you opening the app. They don't need you fluent.

You earned 150 XP today. Your German didn't move.

Not a design flaw — a design choice

XP was invented for Pac-Man. It measures one thing: whether you came back. The app has no idea if you said the sentence correctly — only that you tapped the screen. You were being rewarded for attendance. Never for progress.

Every Monday, you start from zero.

Anderson, ACT-R (1993)

Tuesday's lesson has no memory of Monday. No connection between your pronunciation failure and your vocabulary gap. No compound effect. A coach who forgot everything you told them — every single week. That's not a learning system. That's a content library with a timer.

The app teaching you French is teaching it to 50 million people the same way.

Skehan (1998) — individual differences

None of them have your accent. Your gaps. Your dialect target. Your specific comprehension ceiling. The app built one path and sent everyone down it — beginner or advanced, Parisian or Québécois. You got a mass product dressed up as personal coaching.

You said "eau." The app said "great!" A native speaker would have said "what?"

Flege (1995) — Speech Learning Model

Speaking exercises in every major app are theatre. Your microphone picks up audio. The app plays back a native recording. It has no idea whether your vowels matched — it was never trying to find out. The "well done" was scripted before you opened your mouth.

34% better at reading. Zero better at speaking. Duolingo's own data.

Duolingo internal study, 2022

Duolingo's research team published this in 2022 — not a critic, their own people. Measurable reading gains. Virtually none in speaking. They didn't change the product. The streak was doing its job: keeping you opening the app. Your fluency was never the metric.

Not for the test. Not for the interview. Not for the conversation you've been waiting to have.

CEFR Companion Volume (2020)

Every major app maps XP to CEFR levels without measuring a single CEFR criterion. You were told you were B2. You found out otherwise when it mattered. That's not an accident — the app optimised for lesson completion. Real-world performance under pressure was never part of the plan.

They built for the streak. We built for the conversation.

THE BREAKTHROUGH

Nobody told you fluency
is a physical skill.
That changes everything.

The moment you understand that fluency is physical — built through the same mechanisms as muscle memory, requiring the same honest feedback loop, compounding the same way — every language app you've ever used suddenly makes sense. They were measuring the wrong thing. All of them.

Voicely trains fluency the way a serious coach trains athletes: four measurable dimensions, tracked with precision, compounded deliberately over time. Not lessons. Reps. Not XP. Scores that only move when something real changes.

"In a gym, the weight doesn't lie. It either moves or it doesn't. Language needs that same honesty. We built it."

🏋️The Gym
Cardio78%
Strength54%
Flexibility66%
Endurance82%
🗣️Language
Pronunciation74%
Production52%
Comprehension68%
Retention81%

Same science. Different muscles. Both honest.

FLUENCY POINTS™

XP was designed for video games.
Every point you've ever earned was a lie.

XP rewards presence. Click a button, get a point — the app doesn't know if you got anything right, only that you tapped the screen. Fluency Points are different: they only accumulate when a ring improves. No ring improvement, no FP. The number can't lie to you.

XP — The old way

Complete lesson+50 XP ✦
What triggered itYou opened the app
What was measuredYour attendance
What improved—
Session tomorrowStarts from scratch

FP™ — Fluency Points

74
R1
68
R2
52
R3
81
R4
What triggered itMeasurable improvement
What was measuredAccuracy per dimension
What improvedRing 1 +3 · Ring 3 ceiling
Session tomorrowPicks up where Ring 3 stopped

Earned, not given

FP only accumulates when a ring improves. Say the same word perfectly twice — it counts once. The second time has to be harder. Your score compounds only through real improvement, never through repetition.

Four independent scores

Your Pronunciation can be at 74 while Production sits at 52. No aggregate hides where you actually stand. Each ring is its own truth — and you can't fake any of them.

Rings fall, too

If you stop training a dimension, the ring decays. That's not punishment — it's honesty. Fluency maintenance is real work. Voicely tracks it.

Compound intelligence

HEXI reads all four rings after every session and routes the next one toward your weakest dimension. Your Monday becomes Tuesday's starting point. For the first time, your sessions have a memory.

See how FP Rings™ work →

THE SYSTEM

For the first time, you'll know exactly what's broken — and what to do about it.

Not a score. Not a level. Four independent numbers, each measuring a different dimension of real fluency. Your Pronunciation ring at 74 means something specific. Production at 52 means something else. No aggregate hides the truth. Each ring is its own honest answer.

74

Ring 1

Pronunciation

Phoneme accuracy at the dialect level — not whether you sound foreign, but which sounds, and by exactly how many deviations.

68

Ring 2

Comprehension

Active decoding of native speech at full speed. Not translation — processing. Your ear calibrates to the register you are actually targeting.

52

Ring 3

Production

How fast you construct sentences under pressure. The gap between knowing a word and deploying it in a real conversation.

81

Ring 4

Retention

How well structures hold over time. HEXI predicts your personal forgetting curve and resurfaces items before they disappear.

HEXI™ — Compound intelligence

HEXI doesn't ask how you feel. It knows which ring is your ceiling.

Every session, HEXI reads all four rings and asks one question: where is the ceiling, and what does it need? Vocabulary from reading resurfaces in listening. Pronunciation gaps come back in scenarios. What slipped last Tuesday — HEXI tracked it. It's coming back, framed differently, when you need to close it.

HEXI™ IN SESSION

Mid-session. HEXI just noticed something you didn't.

You're not aware of the shift. The next exercise loads seamlessly. But HEXI pulled it because of the pause you left three minutes ago, the phoneme that slipped in your second sentence, and a comprehension gap it's been tracking across your last six sessions. You're about to close a hole you didn't know was open.

Every phoneme scored at the phoneme level — not "close enough."

What you can't hear, you can't produce. HEXI tracks the link and forces the gap closed.

Fluency tracked in real time: pause duration, words per minute, filler rate.

HEXI predicts your forgetting curve before the word fades — and surfaces it before it disappears.

PULSE™ · SCENARIO IMMERSION — CAFÉ MONTRÉAL

HEXI

Bonjour! Qu'est-ce que vous voudriez commander?

YOU

78

Euh… je voudrais un croissant.

HEXI · PRONUNCIATION

/ʁ/ on "voudrais" — scored 52.

That's the phoneme holding Ring 1. Three Island™ reps on this word will shift it.

Flege (1995): L2 phoneme categories form through targeted production practice.

You, todayRing 1 · 62
↓
Your fluent selfRing 1 · 94

Same voice. Same person.
The gap is trainable.

VOICE CLONE

Train against the version of you that's already fluent.

Record 30 seconds. Voicely builds a model of you speaking your target language at native fluency — your voice, your cadence, your timbre. Every session, you hear it back. Not a stranger's voice telling you how to sound. Yours.

Your brain processes the gap between where you are and a native speaker differently than the gap between where you are and yourself at fluency. The second gap is smaller. More achievable. More motivating. That's not theory — it's how motor learning works.

Your voice model Built from 30 seconds of speech. Unique to you.
Pronunciation target You hear yourself hitting every phoneme correctly.
Compound memory HEXI routes your weakest phonemes to your model first.
Hear how it works →

DIALECT SPECIFICITY

Québécois is not Parisian.
Colombian is not Castilian.

Every other app gives you "French." Then you land in Montréal and find out you've been training for the wrong city. Voicely scores you against the exact dialect you're targeting — phoneme deviations, lexical register, and cultural fluency specific to your region. Not generic French. Québécois French. Not Castilian. Colombian.

Québécois French
Parisian French
Argentine Spanish
Colombian Spanish
Brazilian Portuguese
Tokyo Japanese

VOICECAST™

Not a level.
A fingerprint.

Nobody else will ever have your exact four numbers. Because nobody else is you.

Duolingo says B2. That tells you nothing. VOICECAST gives you four numbers — one per ring. Ring 1: 74. Ring 2: 68. Ring 3: 52. Ring 4: 81. Ring 3 at 52 while Ring 4 sits at 81 tells HEXI exactly where your ceiling is — and exactly what's next.

This isn't a score out of 100. It's a map of where you actually are — and where one more month of real training takes you.

01

HEXI appears

"Your personalized French fluency score."

02

Your baseline recorded

Your recorded phrase — this is your starting point.

03

Four rings reveal

Each ring fills one by one. The gap becomes visible.

04

Your next action

Specific. Ring-level. Not "keep practising" — exactly what to do.

500

founder cohort spots

Lifetime pricing · first access

33

dialects modeled

Québécois ≠ Parisian. Each scored separately.

22

workout modes

Phoneme drills to live AI conversation.

1,400 people decided the streak wasn't enough.
500 spots to go live first.

Founder pricing locked in forever. First access to every new feature. Your VOICECAST™ score — narrated in your own voice — before anyone else gets theirs.

Claim your spot →

When you're ready to stop grinding lessons and start training — one plan, one price, 14 days free. No locked features, no fake levels, no upsells. Ever.

PRICING

One clear price.
No tricks, no upsells.

14-day free trial on every plan. Pro from $59/mo.

✓All 4 FP Rings™ included
✓33 dialects
✓HEXI compound AI
✓2 months free on annual

Yourfluentselfisalreadyinthere.

You already know you want to be fluent. You've known for years. Two minutes to find out exactly where you stand — and to start doing something about it.

The streak was never for you. This is.

What's your real fluency level? →

Free · No account required · Join 1,400+ people who decided the streak wasn't enough

Already training? Sign in →