Gemini Sync

Didactical project by M. Marchese, M. Sergio - Undergraduate course in Communication Design - IED Roma

Academic year

2025/2026

An AI-based system that improves the accuracy of subtitles in online videos, making educational content more accessible for deaf and hard-of-hearing users.

Gemini Sync is a project developed by students Marianna Marchese and Matilde Sergio that investigates the role of artificial intelligence in improving the accessibility of video content.

The research behind the project highlights that 74% of young people use online videos as a learning tool, while deaf and hard-of-hearing users rely almost entirely on automatic captions to access this content. However, current captioning systems generate on average 7.7 errors per minute, turning a ten-minute educational video into a fragmented and often frustrating experience.

To address this issue, the project introduces an approach that prioritizes accuracy over speed. Gemini Sync implements an intentional 1.5-second delay in subtitle generation, creating a short buffer that allows Google Gemini’s AI to analyze not only the audio but also the visual and semantic context of the video before producing the captions.

This “accuracy buffer” significantly improves caption quality, increasing precision up to 99%. The interface also includes a comprehension support system: by selecting a word or unclear passage, users can access an instant contextual explanation tailored to their level of understanding.

The project reflects on the relationship between time, technology, and accessibility, showing how even a minimal delay can radically improve the way digital content is experienced.

Gemini Sync was selected as a finalist in the “Experiential & Immersive” category of the international contest The Young Ones - The One Show, receiving a Merit.