Home > Product > Features > MXT1



Preview MXT-1

Trained for news, sports & entertainment use cases

Thanks to AI Transformer technology, MXT-1 is trained on hundreds of thousands of hours of news, sports and entertainment audiovisual content. This makes it perfect for describing and searching content in the broadcast and sports media industries.

Scans more than 500 hours
of video per minute

MXT-1 is designed to work at scale. Whether you need to retrieve an archive collection, or automatically index your daily incoming flux of rushes and streams. MXT-1 will save you a tremendous amount of time, so you can start working on, sharing and selling your content right away.

For all your media assets: video, live streams, photos & audio

MXT-1 understands, describes, summarizes and ultimately indexes all your audiovisual content to make them searchable in seconds.

Cutting costs with energy efficiency

MXT-1 is 7 times more cost efficient than other unimodal AI on the market. We want our customers to be able to use MXT-1 at scale, so a big R&D focus is on reducing the energy consumption of our pipelines, and ultimately, cutting the cost of AI indexing for you.

Jumble of tags no more.

Let’s be honest, if you’ve tried traditional off-the-shelf AIs in the past, you know perfectly well how much noise they can create.

Here’s what you get with MXT-1 vs a traditional provider.





Vision & STT
Video Indexer
Scene description in natural language
Combines modalities for greater accuracy (multimodal AI)
User Interface to control and manage AI outputs
Semantic search
Transcription and translation
Transcription summary
Shot type and segment detection
Segment only
Face and speaker detection
Face only
Face only
Face only
Logo detection
Yes (closed list)
Landmark detection
Monuments only
Object detection
Text detection (OCR)
Analyze original metadata (date, location..)
Links wikidata universal identifier
Face only
External data feed enrichment
Thesaurus Management

You work at Google, AWS or Microsoft and want to help us keep this comparison up to date?
Reach out to us here.

Your archiving super power

Your archiving super power

AI training made easy

We detect over half a million public figures, and you can easily train MXT-1 to recognize your key people with our thesaurus feature. Our "no-code" approach doesn't stop with AI training. We allow you to easily create indexing routines by setting multimodal rules that match your specific needs.

Automatically summarize context

Newsbridge automatically transcribes and summarizes your videos in 100+ languages. Our Speech-To-Text technology indexes your media assets, making them completely searchable within your database (simply search for spoken words or key phrases). Neural network translation is also available, so you can search in your mother tongue and retrieve results for specific media in another language.

Check the supported languages→

Reach indexing consistency

Deeper learning leads to more accurate results. Here at Newsbridge we understand that our AI needs to meet your custom requirements, which is why we have a built-in knowledge graph for you and your team to self-train (by identifying key people, objects or actions). Simply import and sync your existing thesaurus to extend your taxonomy and collective intelligence, using a user-friendly interface.

Learn more →

Forget manual logging

Powerful and high-quality automatic indexing requires a multimodal approach. As our underlying technology at Newsbridge, MXT-1 helps eliminate false positives by cross-analyzing all media sources (audio, images, metadata…), meaning your team can focus on more specialized editorial tasks. MXT-1 describes your content in a factual and objective manner and is constantly improved to fight bias.

Tell me more →

Generative and Multimodal AI 
(video-to-text and metadata)

MXT-1 is a generative AI that is multimodal, meaning it detects various elements in the frames and audio, and will use everything it has detected to generate a comprehensive and factual annotation, just like a professional media logger would.

Your dataset is private
and always will be.

MXT-1 is not trained by merging our customers’ data. Instead we believe in highly specialized datasets. This is why your dataset remains yours, and yours alone. If you choose to train MXT-1 on your data, we can specifically deploy fine-tuned MXT-1 models for you.

Portable & multilingual metadata

All the metadata extracted by MXT-1 can be sent to a third party library or OTT platform to boost the discoverability of your content.

The metadata can be generated in several languages, because if your organization is bilingual, then your metadata should be too.

“MXT-1 is aligning video perception with the power of language models."

Fred Petitpont, CTO & Co-founder, Newsbridge

Our range of integrations make Newsbridge easy to adopt.
Check our integrations

At the forefront of video search

Newsbridge's mission is to enable you to find the right audiovisual content in less than 2 seconds. Our R&D and product teams are dedicated to pushing the boundaries of AI and search for video use cases. When you choose Newsbridge, you benefit from our continuous innovation.

Your success matters to us

When adopting Newsbridge, a dedicated customer success manager will help you get the most out of your platform. We take pride in our responsiveness, and you can easily reach us via an embedded chat or WhatsApp.

DON't be shy

Get in touch with our team today
We'll give you a demo and a 7-day free trial

Let's go

More features


Meet us November 14 - 16
ADNEC, Abu Dhabi – UAE
Book a time to meet
and demo Newsbridge:
Let's meet


Meet us 24-26 October 2023
Javits Center
New York
Book a time to meet
and demo Newsbridge:
Let's meet