Assignment veiw
Log in page

Wordy Birdy

Inspiration

During the pandemic, reading achievement across U.S. schools dropped by 17%, and the impact was even worse for students in low-income communities.
We saw firsthand, working in classrooms, how many bright, curious kids struggled simply because one-on-one support wasn’t possible.
Wordy Birdy was born from that experience, the belief that every student, no matter their background or learning difference, deserves a patient reading companion that listens, supports, and encourages.

What it does

Wordy Birdy is an AI-powered reading coach that helps students build reading fluency and confidence while giving teachers tools to track progress.

Teachers can upload PDF reading assignments and monitor student performance through accuracy scores and reading analytics.
Students can read aloud, receive instant feedback on pronunciation and fluency, and listen to the correct pronunciation using text-to-speech.
The app tracks misread words, calculates reading accuracy, and offers personalized encouragement and comprehension questions through GPT.

How we built it

Backend: Flask (Python) for the API and routing logic.
Database: Supabase (PostgreSQL) to store assignments and reading submissions.
AI Models:
- Whisper for speech-to-text transcription.
- GPT-4o-mini for personalized feedback and comprehension.
- OpenAI TTS for voice playback and pronunciation.
Frontend: HTML, CSS, and JavaScript for a lightweight, accessible design suitable for classrooms with limited technology.
Libraries: PyPDF2 for text extraction, difflib for accuracy comparison, httpx for API requests, and dotenv for environment management.

Challenges we ran into

Audio processing latency: Whisper transcription initially caused delays; we optimized it by splitting audio files into smaller chunks.
PDF extraction: Non-standard formatting caused text recognition issues, requiring normalization and error handling.
Tone tuning: Ensuring GPT feedback was age-appropriate and encouraging, not robotic or critical.

Accomplishments that we're proud of

Built a fully functional prototype in under 24 hours that integrates speech, text, and AI feedback seamlessly.
Created a clean, accessible design that can be used by both teachers and students.
Seeing the demo in action and hearing the AI coach give warm, personalized feedback made the project come alive.
Stayed true to our mission of accessibility through empathy, which is deeply personal to our team.

What we learned

How to combine multiple AI models (Whisper, GPT, and TTS) into a single coherent user experience.
The importance of designing for inclusion first; when you build for accessibility, you end up helping everyone.
Empathy and technology can coexist; the best educational tools feel human, not mechanical.
We learned how to deploy a Flask app with Supabase and handle live audio data streams efficiently.

What's next for Wordy Birdy

Gamification: Add badges, reading streaks, and progress rewards to motivate students.
Multilingual support: Expand to ESL learners and language acquisition classrooms.
Teacher analytics: Long-term data visualization to show reading growth over time.
Mobile app version: To make it accessible beyond the classroom and usable offline.
Integration with special education tools: Partner with accessibility programs to bring reading confidence to more students.