Inspiration

Active immersion for language learning

What it does

Gives language learners a recurring, active way to practice the language they are learning by sending them updates from various AI character chatbots they can interact with.

How we built it

Vibes with a side of human labor, taste, & debugging.

Challenges we ran into

This project was meant to be part of a larger language learning framework to be integrated with an AI-powered music tool for cross-lingual singing voice synthesis (converting a singer's voice in a song to another language while maintaining identical prosody to the original song). The motivation for the music tool is that many listeners would be familiar with the song lyrics in English, which would facilitate understanding the song in another language. Additionally, it would serve as another medium of language learning. However, current models are not generalizable enough for any given artist. A foreseeable challenge with our text messenger app is how engaging and unique the AI foreign characters are with the user.

What we learned

Cross-lingual singing voice synthesis is difficult and a relatively unexplored topic in audio-language modeling.

Share this project:

Updates