Inspiration

Many students and job seekers struggle with speaking confidently during interviews. Most preparation tools focus on theory, not real speaking practice.
This project was inspired by the need for a platform that allows users to actually speak, get analyzed, and receive actionable feedback on their communication.


What it does

This is an AI-powered interview practice and communication improvement platform.

Users can start a live session where the system:

  • Listens to speech in real time
  • Generates a transcript
  • Tracks words per minute (WPM)
  • Detects filler words like “um” and “uh”
  • Calculates a confidence score

The platform then uses AI to provide personalized feedback to improve clarity, pacing, and delivery.
After each session, a report is generated with insights and improvement suggestions.


How I built it

  • Frontend built using React, TypeScript, Tailwind CSS
  • UI components using shadcn-ui
  • Speech captured using Browser Web Speech API
  • Real-time analytics for WPM, filler detection, and confidence scoring
  • AI feedback powered by Gemini API
  • Session handling and live pipeline architecture for transcript → analytics → feedback

Challenges I ran into

  • Handling real-time speech transcription reliably
  • Syncing transcript, analytics, and AI feedback together
  • Managing microphone permissions and browser limitations
  • Preventing feedback delays and API overload
  • Designing a smooth live-session experience

What I learned

  • Integrating speech recognition into web applications
  • Building real-time analytics pipelines
  • Using AI APIs for feedback and coaching
  • Managing live session state and performance
  • Improving UX for interactive AI tools

Accomplishments that I'm proud of

  • Built a working real-time speaking practice platform
  • Implemented live transcript + analytics pipeline
  • Integrated AI coaching feedback into the session flow
  • Designed a full session → report workflow

Built With

Share this project:

Updates