VocabVoice: A Voice Agent That Helps You Learn and Speak New Words

The Problem
Vocabulary apps teach people to recognise words, not use them. Learners read flashcards but freeze when they have to speak. The gap is active recall and spoken practice, and most apps don’t address it.

The Insight
Words stick when learners hear them, say them and use them in context. So instead of building one more flashcard app, I designed a voice-first loop: Listen → Speak → Use in a sentence.

What I Built
VocabVoice is a working voice-agent prototype that acts like a friendly vocabulary tutor. It explains a word, listens to the learner say it, checks it, then asks them to make their own sentence and gives instant feedback.

Key Product Decisions

  • Zero-cost, zero-setup MVP: I used browser-native Web Speech APIs, so there is no backend, API key or paid service. Anyone can open the link and try it in seconds, with no signup.
  • Honest feedback: I labelled the check “Quick pronunciation check” instead of claiming AI pronunciation scoring. This sets accurate expectations and builds trust.
  • Retention loop: Streaks, a Daily Challenge (3 words) and an accuracy tracker give learners a reason to come back every day.
  • Graceful failure: If the mic is blocked or unsupported, the app still works via Listen and guided practice, so no user hits a dead end.
  • Personalised difficulty: Beginner, Intermediate and Advanced filters let learners practise at their own level.

Scope & Execution
I defined the user journey and MVP scope, designed the UI and edge-case states (listening, no speech, mic denied), built it with AI tools and deployed it free on GitHub Pages.

What I’d Build Next

  1. Real pronunciation scoring
  2. Spaced repetition based on weak words
  3. Adaptive difficulty from the learner’s accuracy
  4. Metrics to track: Day-7 retention, words learned per user, sentence-practice completion rate

Tech Stack: HTML, CSS, JavaScript, Web Speech API, localStorage, GitHub Pages

29 Oct 2026

Keywords
Web Speech API
EdTech
Product Management
GitHub Pages
Voice Agent
JavaScript