Adding on to this hackathon project that lets you use your head to control Chinese flashcards.
Twenty years ago, I was learning Chinese in a small village in central China, and every day I’d walk past signs, menus, and newspapers filled with characters I wanted to learn.
But there was no easy way for me to capture them and turn them into study materials.
So I’ve been building the tool I wish I had back then!
It’s a flashcard app, but watch the demo…
Snap a photo of any Chinese text 📸
NVIDIA micro-services process it quickly ⚡
Select the characters you want to study ✅
Save them directly to your deck 💾
Just point, save, and learn. I’ve been testing it everywhere: Restaurant menus, my son’s Chinese books during story time, a fortune from a restaurant vending machine…
The beauty of this is you’re not learning from some textbook’s idea of useful vocabulary. You’re learning from YOUR world. The signs you pass, the food you order, the books you read. Stuff you will see every day to reinforce your learning.
It was so easy to plug in NIM (NVIDIA Inference Micro-services) into my project. NIMs are containerized services that run on NVIDIA GPUs, exposing generative AI models as APIs.
I just swapped out the vision model by copy/pasting in a new model name into the API call "model": "nvidia/nemotron-nano-12b-v2-vl", commit and push up the new line of code and viola I can upgrade that easy.
Here is the flashcard portion of the app, I still need to adjust the neck movement tolerance so its not so neck-breaking.