"Sottovoce - N DIVIJ team on AMD Developer Hackathon: ACT III Hackathon"

Team Idea

I'm building Sottovoce: AI audio description that lets blind and low-vision children follow lesson videos. The lesson waits for the child, a second model checks every line, and the child can stop and ask. Built solo on AMD Token Factory (Gemma 4, Qwen3.8), Google ADK and Evolus. Live at sottovoce-ai.web.app.

Team Members

N DIVIJ N DIVIJ

div75

AI Engineer

Submission

Sottovoce: lessons a blind child can follow

In two NASA lessons, 24 of 32 descriptions a blind child needed did not fit the teacher's pauses. Sottovoce makes the lesson wait, describes the screen with Gemma 4 on AMD, checks every claim with a second model, and lets the child ask.