# Innovian

- **Event:** [Meta Global AI Developer Hackathon](https://cerebralvalley.ai/e/meta-ai-developer-hackathon)
- **When:** Oct 26, 12:00 PM – Nov 4, 9:00 PM (EST)
- **Where:** Online
- **Team:** [Whit Waldo](https://cerebralvalley.ai/u/whitwaldo)
- **GitHub:** https://github.com/WhitWaldo/lingual-demo
- **Demo video:** https://youtu.be/vE2Xrk2zTvg
- **Gallery:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4405

Following a family conversation is hard with hearing loss. Hearing aids amplify every voice at once, so a dinner table becomes a single wall of sound. It's difficult to tell who said what, and by the time you've pieced together one sentence, the conversation has moved on. Many people end up nodding along instead of taking part.

Lingual listens to the room and turns the conversation into a timeline where each person has their own color and their words appear within seconds. There's nothing to set up: open the app and it's already listening. It recognizes people by their voices, remembers anyone you choose to name so they're labeled automatically next time, and notes the tone each line was spoken in. Tap one or two people to focus on their conversation while Lingual keeps listening to everyone else. Anything said in another language is translated into your own, with the speaker's original words shown underneath.

The hard part is a real room. People interrupt, answer without pausing, and talk over background chatter. Lingual detects when one voice hands off to another even with no gap between them, and it ignores conversations across the room. When crosstalk makes a line impossible to understand, Lingual marks it as unclear instead of inventing words.

Privacy is part of the design, not an add-on. Lingual is designed to run locally on a phone, with nothing uploaded. Only the last two minutes of conversation exist, in memory, and nothing is recorded. The only thing saved is the voice print of someone you've chosen to name, with their permission, and you can delete it at any time.

The demo was executed and recorded on a desktop PC. Future iterations would run the full pipeline on an actual phone and re-speaking each person's words in a clean copy of their own voice, delivered straight to the user's hearing aids, so they hear one clear voice at a time instead of the whole room.

## More from Meta Global AI Developer Hackathon

- [jaguar](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4402)
- [The Magic Table](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4403)
- [Zenith](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4404)
- [Stardive](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4406)
- [Ishan's Team](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4407)
- [David Castro - Agent Guard](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4408)

---

Markdown version of https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4405. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt