AccentGuard

Empowering diverse voices in professional settings through AI-powered communication coaching.

Comment

Inspiration

AccentGuard was born out of a simple but powerful insight: communication confidence should not depend on your accent. Talented professionals feel their ideas aren’t heard because their accent, pace, or filler words hold them back in high-stakes situations Many of us—whether in interviews, presentations, or team meetings—have experienced moments where we worried more about how we sounded than what we said. We wanted to build a tool that empowers people to focus on ideas, not insecurities.

What it does

AccentGuard is an AI-powered communication coach that provides real-time feedback on your speech. Users can:

  • Choose a practice mode: Interview, Presentation, or Meeting
  • Speak naturally while the app transcribes in real time
  • Receive specific, actionable feedback (e.g., “You used um 3 times, try pausing silently instead”)
  • Use built-in demo phrases to test scenarios quickly

The goal is to turn nervous speech into confident communication.

How we built it

  • Frontend: HTML5, CSS3, and JavaScript.
  • Speech recognition: Web Speech API for live transcription.
  • AI feedback: Google’s Gemini API analyzes transcripts with carefully designed prompts.
  • UI/UX: Modern, responsive design with cards, gradients, and hover effects.

Challenges we ran into

  • Learning new tools: This was our first time working with both the Gemini API and the Web Speech API, so we had to quickly understand their uses and limitations.
  • Time pressure: Designing, coding, and polishing a full application within the short timeframe pushed us to prioritize features and cut scope smartly.
  • Prompt engineering: Getting the AI to generate specific coaching instead of generic advice required lots of iteration and fine-tuning.

Accomplishments that we're proud of

We are especially proud that we were able to build a working MVP that actually listens, transcribes, and coaches in real time. In addition, we were able to create a clean and modern UI that feels professional yet approachable with responsive layouts and animations.

What we learned

  • How to integrate new APIs quickly: We gained hands-on experience with both the Gemini API and the Web Speech API, learning their strengths, limitations, and best practices.
  • Effective prompt engineering: We discovered that the way you frame a request to an AI drastically changes the usefulness of its feedback. Iteration and testing were key.
  • Balancing UX and performance: We learned how important it is to deliver feedback that feels both timely and helpful, without overwhelming users or slowing down the application.
  • Designing for inclusion: We realized how much of a difference it makes to frame AccentGuard as a confidence coach rather than an accent correction tool.

What's next for AccentGuard

  • Deployment: Package and deploy AccentGuard as a live web app so users can practice anywhere, without needing to run code locally.
  • Multi-language support: Extend speech recognition and feedback to multiple languages so non-English speakers can benefit.
  • Progress tracking dashboard: Let users see trends in filler word reduction, confidence growth, and speaking pace over time.
  • Improve feedback: We can further improve the feedback and make it more accurate.
Share this project:

Updates