BridgeSpeak

web interface with varying levels of word difficulty in pronunciation that is easy, medium, hard and random
for learning english(globally useful language)
for learning german(hard language to learn due to its complex pronounciation of words)

Inspiration

Improving pronunciation is hard without constant feedback, and access to tutors or native speakers is often expensive or unavailable , especially in under-resourced communities. This tool started as a personal project to improve my own pronunciation. Realizing its potential, I made it more user-friendly to help others improve too.

What it does

BridgeSpeak uses AI to evaluate your English pronunciation. You speak a sentence, and the tool gives you objective feedback based on how closely your pronunciation matches the correct phonemes. No tutors needed , all that's needed is just your voice and a browser.

How we built it

The tool is built with Python (Flask) on the backend and JavaScript/HTML on the frontend. It uses OpenAI’s Whisper for speech recognition and compares phonemes using Epitran and a rule-based model.

Challenges we ran into

Reducing dependencies for easier setup Making phoneme comparison accurate and fast Browser compatibility issues Managing server usage limits for the online version

Accomplishments that we're proud of

Built a working online tool that gives real-time feedback Got real users and feedback from the community Made it easy to use even for non-technical users

What we learned

How to integrate AI speech tools into web apps The importance of accessibility and user experience How to build tools that scale and adapt for different learners

What's next for BridgeSpeak

Add support for more languages Improve UI/UX and visual feedback Create a mobile-friendly version Partner with educational programs to reach more learners