Clone ELSA Speak: Revenue, Tech Stack & AI Prompts (2026)

Briefing

What is ELSA Speak?

ELSA Speak (English Language Speech Assistant) tackles a problem traditional language apps duck: pronunciation. Where Duolingo teaches you to read and translate, ELSA listens to you speak, scores you phoneme by phoneme, and shows exactly which sounds you got wrong and how to fix your mouth shape. It's positioned as a personal accent coach in your pocket and monetizes through a Pro subscription.

The magic feature is speech assessment: the app records a learner saying a target word or sentence, runs it through a pronunciation-scoring model, and returns a per-sound breakdown - green for the phonemes you nailed, red for the ones a native speaker would flag. That feedback loop, repeated across thousands of micro-lessons, is the entire value proposition. The content around it (lessons, games, conversation practice) is comparatively ordinary.

For a cloner, the pronunciation scoring is the moat and also the part you should not build from scratch - modern speech APIs (Azure Pronunciation Assessment, Speechace) and speech-LLMs now return phoneme-level scores out of the box. That collapses ELSA's hardest piece into an API call, which means the realistic opportunity is a focused one: pronunciation coaching for a specific native language ('English for Brazilian Portuguese speakers'), a specific accent target, a profession (medical English, aviation English), or a different target language entirely.

Who it's for: Adult English learners who can already read and write but want to be understood when they speak - heavily international (Vietnam, Brazil, Japan, India), often professionals preparing for work, interviews or exams. Clone opportunities target one native-language audience, one profession, or one accent target.

Revenue model

How ELSA Speak makes money

$ ELSA Pro subscription: roughly $11.99/month, with quarterly and yearly plans heavily discounted to push annual commitment; the yearly plan is the conversion target.
$ Feature gate on depth: free users get limited daily lessons and shallow feedback, Pro unlocks unlimited practice, full phoneme breakdowns and the personalized study plan.
$ B2B and education: enterprise and school licensing for employees and students learning business English - a lower-churn second revenue line.
$ Assessment and certification add-ons: speaking-test prep and proficiency reports learners (or their employers) will pay extra for.

Revenue estimate

$1M–$3M/mo

Rough estimate of app-store consumer spend based on public third-party reports; excludes B2B and education licensing. CloneMRR is not affiliated with ELSA Speak; figures are for educational purposes.

Sources:

Spec sheet

Features to build

MVP ship this first

✓ Lesson library

Short pronunciation drills grouped by sound, topic and difficulty, each with a target word or sentence to say.
✓ Record + score

Tap to record yourself, send audio to a pronunciation-assessment API, get an overall score plus per-word feedback - the core loop.
✓ Phoneme feedback

Color-coded breakdown of the utterance: which sounds were correct, which were off, with a tip on how to fix each flagged sound.
✓ Onboarding placement

A short speaking assessment that sets the learner's starting level and targets their weakest sounds - flowing into the paywall.
✓ Paywall + subscriptions

Free daily lesson limit and shallow feedback; full feedback and unlimited practice behind a subscription with a free trial (RevenueCat / Stripe).
✓ Streaks & progress

Daily streak, score history per sound, and a simple 'sounds you've mastered' view to keep learners coming back.

~ 2–4 weeks with AI builders

Full version add later

+ Personalized study plan

Adaptive curriculum that prioritizes the learner's worst phonemes and revisits them with spaced repetition.
+ Conversation practice

Free-speech prompts and dialogues scored for pronunciation and fluency, optionally with an LLM conversation partner.
+ Accent targeting

Choose a target accent (US/UK/Australian) and score against it; surface the specific sounds that differ from the learner's native language.
+ Exam & professional tracks

Modules for IELTS/TOEFL speaking, business English, or a specific profession.
+ Native-language hints

Tips written in the learner's first language explaining why a sound is hard for them specifically.
+ Offline lessons

Download a lesson set for practice without a connection; queue recordings for scoring when back online.

~ 3–4 months

Architecture

Recommended tech stack

Layer	Our pick	Why
Mobile app	React Native (Expo) or Swift	Microphone capture, audio recording and in-app purchases work well in Expo; pick Swift for the lowest-latency native audio if recording quality becomes a bottleneck.
Pronunciation scoring	Azure Pronunciation Assessment or Speechace API	Phoneme-level speech scoring is the moat - and a solved API problem. These return per-word and per-phoneme accuracy so you don't train a speech model; it's a third-party dependency, own it consciously.
Backend & data	Supabase	Lessons, attempts and progress are simple Postgres rows; proxy the scoring API through Edge Functions so keys never ship in the app and you can cache/meter usage.
Audio storage	Cloudflare R2 + CDN	Reference audio and recorded attempts are the bandwidth cost; R2's zero egress fees matter once practice volume grows.
Subscriptions	RevenueCat	Wraps StoreKit/Play Billing, free trials and paywall A/B testing without writing receipt-validation code.
Analytics	Amplitude or PostHog	Lessons-per-day, score improvement over time and trial-to-paid conversion are the numbers that decide whether learners stick and pay.

The payload

AI prompts to clone ELSA Speak

Pick your builder, copy the prompt, paste it and iterate. Enter your email once to unlock all prompts on every page - we'll also send you this full prompt pack.

elsa-lovable.md

Build an English-pronunciation coaching web app called SpeakSharp, modeled on ELSA Speak.

## Core concept
Learners record themselves saying target words and sentences; the app scores their pronunciation and shows exactly which sounds were off and how to fix them. A few lessons are free; full feedback and unlimited practice sit behind a subscription with a free trial.

## Pages
1. Landing page: warm hero with a mic graphic and a sample score card ('You said "three" - 82%'), headline 'Speak English clearly. Get coached on every sound.', email signup, pricing (monthly $9.99 / yearly $59.99 highlighted), FAQ
2. Placement test (after signup): 5 short sentences to read aloud; produces a starting level and a list of weak sounds, ending on a paywall with free-trial offer
3. Home: greeting, daily streak pill, 'recommended for you' lessons targeting weak sounds, category rows (Sounds, Sentences, Conversation), lock icons on premium lessons
4. Practice screen: the target phrase displayed, a big record button, then a results card - overall score, the phrase with each word colored green/amber/red, and a tip per flagged sound; 'try again' button
5. Progress: streak calendar, score trend chart, a 'sounds mastered vs. needs work' grid
6. Admin area (role-protected): CRUD for lessons - phrase, target sounds, category, difficulty, is_free toggle

## Accent & native language
Let the learner pick their native language and a target accent (US/UK) at onboarding; use those to choose which sounds to emphasize and how to phrase tips.

## Data model
- users: id, name, email, native_language, target_accent, is_premium, trial_ends_at, streak_count
- lessons: id, phrase, target_phonemes, category (sound/sentence/conversation), difficulty, is_free
- attempts: id, user_id, lesson_id, audio_url, overall_score, per_word_json, created_at
- sound_progress: user_id, phoneme, best_score, attempts_count

## Business rules
- Free users get a small daily lesson allowance and only an overall score; Pro unlocks the full per-sound breakdown and unlimited practice
- For scoring, call an external endpoint (stub it with a mock that returns a realistic overall score and per-word colors/tips - structure the code so a real pronunciation-assessment API drops in)
- Record audio in the browser (MediaRecorder), upload it, then request a score
- Trial: 7 days of full access via Stripe Checkout subscription with a trial period; show days remaining in the header

## Design
Encouraging and clean: friendly off-white background, a confident teal/coral accent, rounded-2xl cards, soft shadows, clear sans (Inter). The score card is the hero element - make green/amber/red feedback feel motivating, never punishing. Mobile-first.

Use Supabase for auth, database and audio storage, and Stripe for subscriptions in test mode. Seed with 12 lessons across the 3 categories with realistic phrases and target sounds, and one demo user with a streak and a few past attempts.

Locked

Unlock the full prompt

Free - enter your email and we'll unlock all 5 prompts site-wide and send you the complete ELSA Speak prompt pack.