Sentence Builder
Arrange words into useful sentences and learn the pattern, not only the answer.
Your life is the curriculum
Learn the language you actually need to speak.
VoiceIsland turns your routines, relationships and real conversations into personal speaking practice, then helps you hear it, say it and remember it.
Any language pair · Built around your own words · Free invited beta
One practical learning loop
From the words in your day to sentences you can use.
Three focused stages turn personal context into spoken language. No generic tourist script. No isolated vocabulary treadmill.
Stage 1 · Context
Start with what you already say.
Talk, type, paste or import moments from your real life. The AI coach asks useful follow-ups, cleans your thoughts and creates natural bilingual cards.
- Guided conversations and free narration
- Personal sentence cleanup and translation
- Review every card before it enters practice
Stage 2 · Shadowing
Train your ear, rhythm and voice together.
Hear a natural model, repeat it immediately, record yourself and compare both versions. Slow it down, loop it and stay with one sentence until it feels physical.
- Natural audio with speed and loop controls
- Record, replay and compare pronunciation
- Word-level speech assessment and feedback
Stage 3 · Active Recall
Make the sentence available when it matters.
See the meaning in your strongest language and produce the target sentence without looking. VoiceIsland schedules the next review from your real memory performance.
- Speak from memory before revealing the answer
- Adaptive FSRS repetition instead of fixed drills
- Clear progress across cards, words and pronunciation
Built for daily use
More ways to turn practice into speaking.
Starter Islands
Begin with a practical foundation before your personal library is ready.
14-day sprint
Follow one structured speaking plan with daily tasks and benchmarks.
Pronunciation
See what the app heard and which words need another attempt.
Today
Open the highest-value session based on your real learning history.
Driving mode
Use your existing cards in a voice-led, hands-free practice session.
Your reason, your island
Learn for the conversation waiting outside the app.
Explain what you actually do
Prepare updates, meetings, introductions, questions, and the situations specific to your job.
Language learning for work →Go beyond tourist scripts
Practise variations, follow-up answers, and recovery language for the trip you are really taking.
Language learning for travel →Handle the conversations courses miss
Build islands for housing, healthcare, neighbours, administration, and belonging.
Language learning for expats →Speak in the moments that matter
Prepare school conversations, routines, feelings, plans, and family stories.
Language learning for parents →Build in public roadmap
A clear path from working prototype to a language-learning product.
Every milestone reflects the real application today. Filled markers are live; empty markers are still ahead.
37
of 59 milestones complete
63%
Completed
Next priority
Planned
Product foundation
The technical base required for a usable cross-device prototype.
Web application and Vercel deployment
Responsive product experience deployed on the production domain.
Landing page and installable PWA
Product website, mobile layout, manifest and offline application shell.
Email and Google accounts
Protected user access and session management through Clerk.
Neon cloud synchronization
Cards, conversations and progress available on phone and computer.
OpenRouter AI connection
Project-funded text generation, translation and speech generation.
Material capture
Collect the language a learner genuinely needs in daily life.
Any source and target language
Language pair, practical topics and custom conversation subjects.
AI interview coach
Focused questions and contextual follow-ups in the learner's strongest language.
Typing, pasting and browser dictation
Fast capture without forcing the learner into one input method.
Conversation file import
TXT, Markdown, CSV, WhatsApp-style exports and LLM conversations.
Persistent conversation sessions
Chat-style history, saved transcripts, resumable sessions and per-conversation card progress.
Uploaded audio transcription
Provider-independent speech-to-text for voice notes and recordings.
Realtime voice conversation
Natural spoken dialogue with an AI coach instead of turn-based text only.
Language Islands
Turn rough personal input into clean, reusable practice material.
AI sentence cleanup and translation
Natural source sentences and conversational target-language versions.
Personal card and story generation
Reusable sentence cards plus a bilingual mini-story from the same material.
Human review before approval
Edit, select or reject every AI-generated sentence before it enters practice.
Sentence library
Bilingual cards, notes, topics, audio controls and removal.
Duplicate prevention
Existing material is checked before new cards are accepted.
Prompt evaluation suite
Versioned prompts and repeatable checks for grounding, placeholders, duplicates and card metadata.
Grammar and difficulty intelligence
Automatic grammar tags, CEFR approximation and a 1-5 complexity score on every card.
Audio and shadowing
Build listening familiarity, rhythm and confident physical production.
Natural OpenRouter speech
AI-generated MP3 with a device voice fallback when required.
Shadowing player
Single sentence, previous/next, full playlist and continuous looping.
Persistent Vercel Blob storage
Generated audio remains available across sessions and devices.
Global audio deduplication
Matching text, model and voice reuse one stored recording.
CDN cache and retention policy
Fast playback with cleanup after 180 days without use.
Pronunciation recording loop
Listen to the model, record yourself and replay both versions.
Pronunciation assessment
Transcription, word-level differences, rhythm and pacing feedback.
Recall and measurement
Convert recognition into speech that is available under pressure.
Active Recall scoring
Produce from memory, compare the answer and receive a 0-3 score.
Adaptive spaced repetition
A clear memory rating schedules each card at the right moment.
Daily 30-minute plan
Conversation, shadowing and due-card practice in one learning loop.
Personalized next-session recommendations
Today selects the highest-value activity from due cards, difficult material, pronunciation, benchmarks and sprint progress.
History and repeatable benchmarks
Sessions, recalls, audio plays, activity and comparable test scores.
AI-generated 14-day sprint
Personalized daily sessions, starter phrases, real activity tracking and three speaking benchmarks for any language pair.
Adaptive FSRS scheduling
Personal intervals driven by memory performance instead of fixed days.
Speaking baseline and weekly trend
Repeat the same five-card test and compare accuracy, active vocabulary, pace and independence.
Pre-Input Comprehension
Import native transcripts, preview key vocabulary and approve useful phrases before watching.
Mobile routine
Make focused practice available during real gaps in the day.
Responsive installable mobile app
PWA layout designed for phone practice and home-screen installation.
Cross-device learning state
Continue the same library and progress record on mobile and desktop.
Practice notifications
Three local routine prompts plus a daily Web Push reminder after the installed PWA is closed.
Hands-free Driving Mode
Voice-led sentence and vocabulary sessions with automatic scoring, spoken feedback and Media Session controls.
Offline audio playlists
Download the selected Shadowing set before commuting or travelling.
Lock-screen audio controls
Pause and move through Shadowing or Driving Mode from supported headphones and lock screens.
Capacitor mobile shell
Android and iOS projects are generated and synchronized from one production configuration.
Commercial system
Control operating costs while keeping onboarding effortless.
Project-funded AI and rate limiting
Learners can use the core product without configuring a provider.
Per-user AI usage ledger
Track text tokens, TTS, transcription, cache reuse and cost by learner, feature and model.
Administrative model controls
Validated runtime selection for text, fallback, TTS and transcription models.
Customer-ready settings
API and OAuth configuration is available only to the product administrator.
Fair-use protection
Monthly budgets, request limits, dashboard warnings and graceful limits.
Stripe subscriptions
Trials, promotion codes, recurring payments, invoices and cancellation.
Plan entitlements
Server-side AI access and monthly allowances for Free, Beta, Learner and Coach.
Privacy and data lifecycle
GDPR controls, account deletion and private treatment of sensitive audio.
Beta and launch
Validate learning outcomes, reliability and commercial readiness.
Automated product test suite
Unit, API and end-to-end coverage for the complete learning workflow.
Monitoring and production alerts
Errors, latency, failed generations, sync health and cost anomalies.
Closed learning beta
10-20 active learners with interviews and retention measurement.
Language quality validation
Human review of important language pairs and model outputs.
1,000-user load test
Validate synchronization, AI limits, Blob delivery and database capacity.
Security, legal and support readiness
Audit, policies, consent language, onboarding and help centre.
CarPlay and Android Auto
Native vehicle templates, steering-wheel controls, audio focus and physical-car validation.
Native store releases
Signed builds, physical-device validation, Google Play and Apple App Store review.
Brand decision and paid public launch
Final identity, product positioning and commercial release.
Pricing
Start free. Keep the speaking habit when it works.
Founding members help shape the product and keep their early price while their subscription remains active.
Invited beta
Free
For early testers willing to share honest feedback.
- Personal context and starter islands
- Shadowing and active recall
- Limited AI and audio during beta
Founding Learner
£9 / month
For learners building a real daily speaking habit.
- 7-day free trial
- Personal AI learning materials
- Speech, benchmarks and adaptive reviews
- Founding price kept while subscribed
Teams and families
Later
Shared learning without compromising personal practice.
- Shared practice packs
- Partner and family progress
- Tutor-managed learning paths
- Early access waitlist
Start with the sentences already hiding in your day.
Open the prototype, answer one question, and create your first language island set.
Open VoiceIsland