AI Speaking Coach

Practice your speaking skills, get AI feedback, and track your progress — all in the browser.

What it does

• Practice — Pick a random topic, hit record, and speak for up to 2 minutes. The app transcribes your speech in real-time using the Web Speech API.
• Filler word detection — Catches words like "um", "uh", "like", "you know" and highlights them in the transcript.
• AI feedback — After recording, get feedback on clarity, structure, vocabulary, and confidence via OpenRouter (free models work).
• Pronunciation — Embedded ai-pronunciation-trainer for phoneme-level accuracy scoring with color-coded results and IPA comparison.
• Analytics — Track your daily streak, total sessions, speaking time, and browse your history.

Setup

git clone https://github.com/Vivekbhatt02/AI-Speaking-Coach
cd "AI Speaking Coach"
npm install
npm run dev

Opens at http://localhost:5173. Use Chrome or Edge for best speech recognition support.

AI feedback (optional)

1. Grab a free key from OpenRouter
2. Click "Add API Key" on the Practice page
3. Uses nvidia/nemotron-3-nano-30b-a3b:free by default

Pronunciation (optional)

The pronunciation page embeds the hosted aipronunciationtr.com trainer via iframe — no extra setup needed.

Scripts

• npm run dev — dev server (port 5173)
• npm run build — production build
• npm run test — run tests

Built with

React 18, JavaScript, Tailwind CSS, Vite, Vitest, Web Speech API, MediaRecorder API. Data stored in LocalStorage — no backend required.

Project layout

src/
  components/    UI components (TopicGenerator, SpeechRecorder, FeedbackPanel, etc.)
  pages/         Route pages (Home, Practice, Pronunciation, Analytics)
  utils/         Filler word detection, localStorage helpers, streak logic
tests/           Test files (Vitest + React Testing Library)

Troubleshooting

• "Web Speech API not supported" — Use Chrome, Edge, or Safari
• "Microphone access denied" — Allow mic permission in browser settings
• Blank transcript — Check your mic input in system settings
• Firefox — Speech recognition is limited, Chrome/Edge recommended