Inspiration
Public speaking is one of the most common and anxiety-inducing skills people struggle with, yet meaningful feedback is hard to get without a coach/instructor. We wanted to make high-quality speech feedback accessible to anyone with a microphone.
What it does
SpeechMentor is a speech coaching web app. The goal is to help users speak more clearly and confidently.
- A full transcript
- Words-per-minute (pace)
- Filler words
- Confidence and clarity scores
- Actionable feedback on how to improve
How we built it
- Frontend: HTML, Tailwind CSS, and JavaScript using the Web Audio API for in-browser recording
- Backend: Node.js + Express
- Speech-to-Text: ElevenLabs Speech-to-Text API
- Analysis: Currently using custom logic for pace calculation, filler word detection, and scoring. We hope to later integrate with Gemini API for real feedback.
Challenges we ran into
- Handling audio recording and encoding reliably across browsers using the MediaRecorder API
- Correctly formatting multipart audio uploads for the speech-to-text API
Accomplishments that we're proud of
- It sorta works!
What we learned
- Audio handling in web apps is more complex than it looks
- Some of our group learned version control during this hackathon
What's next for SpeechMentor
- Session history and progress tracking
- Tone, sentiment, and structure analysis
- Actual feedback rather than custom hard-coded feedback!
Built With
- elevenlabs
- express.js
- firebase
- html
- javascript
- node.js
- tailwind
- webaudioapi
Log in or sign up for Devpost to join the conversation.