Newsbridge MXT-1 – BaM Award winner, Manage
MXT-1 is a patented, multimodal and generative AI indexing technology by Newsbridge that uses natural language models to generate human-like descriptions of video content. Capable of indexing more than 500 hours of video per minute, MXT-1 is a game changer for organizations working with media and sports audiovisual content. Leveraging the next-gen technology, users can index vast amounts of content in record time, and search through large video collections easily and efficiently, enabling them to quickly start enhancing, sharing, and monetizing their media assets.
MXT-1 offers a unique, multimodal approach to AI indexing. Combining computer vision and speech processing with natural language models, and leveraging AI transformers, the technology is specifically trained on hundreds of thousands of hours of media, entertainment, and sports content. MXT-1 detects people, logos, context, actions, landmarks, and text, and transcribes, translates, and summarizes audio to text. It then merges these data sources to produce a human-like, objective description. While traditional AI providers in the market are only capable of generating keywords and speech-to-text content, MXT-1 uses natural language to link everything together and describe a scene, resulting in more accurate descriptions and faster, more affordable AI indexing.
What’s particularly innovative is MXT-1’s ability to describe video scenes just like a human would. These unique features and capabilities also set MXT-1 apart from its competitors:
- Automated Media Summaries – MXT-1 generates extensive descriptions of media assets. With this feature, users gain a comprehensive understanding of video content without needing to watch the footage in its entirety.
- Media Summaries in Multiple Languages – Scene descriptions and media summaries can be viewed in English, French, Spanish, Italian, German, Portuguese, Japanese, and Arabic — with more languages to come. Most other generative AI technologies only support English.
- Multilanguage Descriptions and Speedy Multilingual Transcription – When a media asset features multiple spoken languages, they can be automatically detected and transcribed using MXT-1. These enhanced capabilities complement MXT-1’s existing robust transcription and neural translation features, available in over 130 languages.
- Facial Recognition and Custom Thesaurus – MXT-1 automatically identifies and tags public figures, sports people, presenters, and can be trained to recognize anyone else relevant to an organization.
- Landmark Detection – Locating video clips of historic structures, monuments, cityscapes, and landscapes is lightning fast with MXT-1. The technology even specifically detects aerial shots.
- Logo and Pattern Identification – MXT-1 detects specific logos and on-screen graphics to help users better sort content, frame narratives and even boost relationships with sponsors and partners.
- Approval Workflows – Users can review, modify and verify the MXT-1 AI-generated output.
- Highlight Identification – Media loggers and catalogers can spotlight the most engaging or exciting moments in a video that has been analyzed by MXT-1.
- Energy Efficiency – MXT-1 is seven times more cost-efficient than mono or unimodal AI systems. Newsbridge achieved this breakthrough by maximizing the use of CPUs over the more power-hungry GPUs, coupled with a smart frame sampling technology.
Newsbridge is used worldwide by TV channels, press agencies, production houses, journalists, editors, archivists, and sports rights holders to boost their production workflow and media ROI.
