medIAenrich: AI for enriching audiovisual content

What is medIAenrich?

medIAenrich performs a multimodal analysis of a program based on video, audio, and subtitles.

This analysis helps build a database of information related to the program. This data makes it easier to understand and index the program.

The information analyzed and collected is extensive: summary, segmentation (chaptering), detection of objects, text, faces, logos, shots, etc.

For a television news program, for example, medIAenrich can detect different scenes, allowing it to distinguish between field reports and studio segments:

medIAenrich goes even further:

On the left, we asked a commercial solution to analyze a clip from France 2’s 8 p.m. news. On the right, we can see that medIAenrich’s analysis is much more comprehensive, accurate, and contextualized.

From Research to Industrial Solutions

In October 2019, France Télévisions partnered with Télécom SudParis to establish a joint research laboratory. Researchers from Télécom SudParis, supported by engineers from France Télévisions, focused their research on multimodal analysis and developed the analysis algorithms.

In a second phase, France Télévisions developers leveraged the lab’s findings to create functional services, marking the transition from the R&D phase to the industrial phase. These AI services are now available to European public media organizations.

medIAenrich opens the door to numerous possibilities for enhancing media across all platforms and preparing the audiovisual world of tomorrow.