TurtleTalk

About TurtleTalk

TurtleTalk is an AI-powered platform for Indigenous language learning, featuring interactive exercises, real-time pronunciation feedback via Google Gemini, and neural text-to-speech. Built for HackHive 2026, currently showcasing Plains Cree with plans to expand across Indigenous languages.

Inspiration

Indigenous languages are disappearing rapidly - UNESCO reports many are critically endangered. We wanted to create technology that supports language revitalization while respecting cultural context, making learning as engaging as Duolingo but designed specifically for Indigenous communities.

What it does

Interactive Lessons: Duolingo-style exercises with translation, pronunciation, and listening comprehension
AI Pronunciation Feedback: Google Gemini analyzes speech and provides scores (0-100) with specific improvement tips
Neural Text-to-Speech: VITS model generates authentic Indigenous language audio
Cultural Content: Stories and lessons within cultural context
Community Features: Forums, events, and translation collaboration
Accessibility: Full screen reader support and keyboard navigation

How we built it

Frontend: React + TypeScript, Material-UI, Web Audio API for recording
Backend: FastAPI (Python), SQLAlchemy, JWT authentication
AI: Google Gemini 1.5 Flash for pronunciation analysis, VITS for TTS, Whisper prototype for speech-to-text
Resources: itwêwina Dictionary, Indigenous Languages Corpora, Cree Language Podcast

Challenges we ran into

Browser audio recording compatibility across different devices
Fine-tuning AI prompts for culturally sensitive pronunciation feedback
Adapting TTS models for Cree phonetics with limited training data
Balancing feedback timing (settled on 4 seconds)
Time constraints prevented full speech-to-text integration

Accomplishments that we're proud of

Real AI-powered pronunciation feedback, not just mock responses
Completed speech-to-text prototype for future conversational features
Polished Duolingo-style UI with smooth interactions
Accessibility built-in from day one
Scalable architecture supporting multiple Indigenous languages

What we learned

Multimodal AI integration and prompt engineering for cultural sensitivity
Browser audio processing and format conversions
Building WITH Indigenous communities, not just FOR them
User experience timing is crucial - small details matter
Sometimes simple solutions (SQLite) beat complex ones for iteration speed

What's next for TurtleTalk

Conversational AI Tutor: Integrate speech-to-text prototype with Whisper for interactive dialogue
Multi-Language: Expand to Ojibwe, Mohawk, Inuktitut, Lakota, Navajo
Mobile Apps: Native iOS/Android with offline mode
Gamification: Leaderboards, achievements, social challenges
Community Partnerships: Work with tribal colleges, cultural centers, and language keepers
Impact Research: Measure effectiveness in language revitalization efforts