Pitch Coach

Inspiration

I go to school in Waterloo, Ontario. We have one of the highest startup densities in the world. But the more hackathons I go to and the more startup founders I talk to, I notice one thing; many struggle with pitch delivery, especially under time pressure or nerves. Hiring a coach is expensive, and practicing alone doesn’t provide constructive feedback. I wanted to solve this problem.

What it does

It records your pitch, analyzes your speech using OpenAI’s Whisper for transcription, GPT-4 for real-time feedback, and uses facial recognition to evaluate expressions like confidence and engagement. In just one minute, you get actionable coaching tips on pace, filler words, and body language—all without needing a human coach.

How we built it

Pitch Coach is a full-stack AI web application that combines: React for the frontend UI, Express + Node.js for the backend API, OpenAI Whisper for transcription, OpenAI GPT-4 for feedback generation, face-api.js for facial expression analysis (happy, neutral, angry), and react-media-recorder for capturing audio and video in-browser.

Challenges we ran into

The biggest challenge I faced was getting consistent and accurate feedback from AI required prompt engineering and fallback logic.

What we learned

We learned how to integrate multiple AI APIs and libraries in a seamless full-stack project. We also learned the art of fine-tuning prompts for GPT-4 to produce more actionable, less generic responses.

What's next for Pitch Coach

I didn't have time to add the functionality, but eye contact scoring would be next on the bucket list. I also want to incorporate gesture recognition at some point. I'd also like to add a save function so users can learn from past pitches.

Built With

• javascript
• open-ai-api