- Demo Video: https://vimeo.com/1097441801?fl=pl&fe=cm
- Devpost: https://devpost.com/software/cardemic
Whenever I’d watch an English-language video, I often stumbled upon unfamiliar words. Stopping the video to check the meaning broke my immersion, and the frustration always lingered.
I imagined a smarter solution: an app that could predict unknown words ahead of time, giving learners a seamless and enjoyable experience. Adding swipe-based interactions, inspired by apps like Tinder, could make vocabulary learning intuitive, addictive, and genuinely fun.
Cardemic transforms YouTube videos into a playful vocabulary-building experience.
Simply paste a link, and the app fetches the transcript, identifies words you likely haven’t learned yet, and generates a card-based game. After a short, engaging session, unfamiliar vocabulary becomes second nature, ensuring uninterrupted video enjoyment.
We built Cardemic using React Native and Expo for a smooth, cross-platform experience.
Firebase powers our authentication and backend seamlessly. The core magic comes from Google Vertex AI, which intelligently analyzes video transcripts, selects challenging vocabulary tailored to each user’s proficiency, and creates convincing distractors.
The hardest part was getting the AI to consistently identify unknown words from YouTube transcripts, then deliver accurate translations and convincing distractors.
Moving quickly with AI meant dealing with numerous edge cases, each requiring careful attention to maintain reliability across any language pair.
We successfully created an app that’s language-independent and intuitive.
Without manually coding rules for each language, Cardemic accurately predicts and teaches vocabulary across numerous language pairs. Seeing our concept become a functional, engaging tool felt incredibly rewarding.
Working with AI accelerates development tremendously, allowing rapid prototyping and experimentation that would otherwise take weeks or even months.
However, this increased speed brings its own complexity, requiring us to continuously identify, manage, and refine numerous edge cases that emerge at nearly every step.
Soon, Cardemic will support vocabulary extraction from webpages and PDFs. We also plan to introduce features like leaderboards and streaks to keep motivation high.
Our goal is to make vocabulary learning feel effortless and addictive, turning any content, from anywhere, into an engaging, swipe-based learning experience.