Advanced Television

Moments Lab launches MXT-2

March 18, 2025

AI video discovery company Moments Lab has unveiled MXT-2, the latest version of its multimodal AI indexing technology that helps producers find the right moments in their video libraries for fast content creation, repurposing and monetisation.

At 1.5 billion images and counting, MXT-2’s training dataset is more than three times larger than its predecessor, enabling the AI to generate even more detailed and accurate humanlike descriptions of video content.

New features include Custom Moments, which are time-coded descriptions of key video moments that are tailored to user needs — such as pinpointing the most compelling scenes in unscripted TV shows to build viral shorts, trailers or best-of compilations.

The new Custom Insights feature enables MXT-2 users to automatically generate texts of any length, including titles, teaser descriptions, and full-length articles. Custom Insights also automates the theme or editorial topic classification of videos.

“We’re proud to say that MXT-2 is the best technology in the world to sequence and describe video content,” said Frederic Petitpont, CTO and co-founder of Moments Lab. “In fact, MXT-2 outperforms Google Research’s Vid2Seq on video sequencing by 47 per cent.”

“Our mission is to remove the roadblocks to better video storytelling by making it simpler to search through and sort ever-growing media libraries,” added Philippe Petitpont, CEO and co-founder of Moments Lab. “MXT-2 provides our users with even deeper editorial insights and video searchability at scale to maximise content creation, reuse, and repurposing, and to deliver tangible ROI.”

MXT-2 is available via the Moments Lab video discovery platform, or it can be integrated into users’ existing tools, workflows and platforms via API.

 

Categories: AI, Articles, Search/Recommendation

Tags: , ,