Newsbridge adds AI focused on generating human-like video descriptions

Newsbridge has unveiled MXT-1, an AI indexing technology designed to generate human-like descriptions of video content using natural language models, capable of indexing more than 500 hours of video per minute.

Frederic Petitpont, co-founder and CTO of Newsbridge, said MXT-1 represents a significant breakthrough in content indexing and discovery. The technology combines multiple AI modalities, including computer vision, speech processing, and natural language models, to produce impressive results.

MXT-1 is specifically trained on a vast range of media, entertainment, and sports audiovisual content, making it highly effective for indexing and search purposes in these industries.

Newsbridge expects MXT-1 to dramatically reduces the cost of using AI at scale, enabling mass indexing of media assets to become a business reality. MXT-1’s reduced energy consumption makes AI indexing seven times more cost-efficient than mono or unimodal AI systems.

One of MXT-1’s key advantages is its ability to describe scenes in natural language. The technology links raw modalities, such as face, text, logo, landmark, object, action, and shot type detection, as well as transcription, to generate a semantic description for enhanced searchability. MXT-1 is a considerable improvement over existing AI indexing solutions, which often produce a disorganized collection of tags that fail to provide the necessary information for content owners.

Currently in beta mode, MXT-1 is being gradually incorporated into all of Newsbridge’s cloud solutions, including Just Index, Media Hub, Media Marketplace, and Live Asset Manager.