Inspiration
We're a team of three — French, Indian, and Austrian — all living in the U.S. We realized that the future of robotics shouldn't require you to speak English. What if you could talk to a robot in your mother tongue and it just... understood? That's how Babel was born: a multilingual voice interface for robots, built by a multilingual team.
What it does
Babel lets you command a Unitree Go2 robot dog using your voice in any language — French, Hindi, German, or English. The robot listens, understands, executes the command, and responds back in English. No buttons, no keyboard. Just talk.
How we built it
We combined Smallest.ai Pulse for multilingual speech-to-text, GPT-4o-mini for intent parsing and translation, Cyberwave SDK for robot control. Everything runs through a single Python pipeline: voice in → text → command → robot moves
Challenges we ran into
Getting reliable multilingual transcription in a noisy hackathon environment. Mapping natural language ("go left", "fais un flip") to actual robot API commands. Syncing three teammates across three languages and one codebase.
Accomplishments that we're proud of
Built a fully working voice-to-robot pipeline in one day. Three languages, one robot, zero buttons. We even got it to dance — which technically isn't in the API, so either we're geniuses or the robot just vibes with us.
What we learned
How to work with real-time STT/TTS APIs under pressure. That robots don't speak French (yet). And that the best teams are the ones where nobody speaks the same language.
What's next for Babel
Adding more languages (Japanese, Arabic, Portuguese). Real-time conversation mode where the robot can ask clarifying questions. And expanding beyond the Go2 to other robot platforms via Cyberwave.
Log in or sign up for Devpost to join the conversation.