Generative AI has unleashed incredible innovation across the broadcast and media industry, and the technology rightfully took center stage at this year’s IBC Show. It was a special show for Newsbridge as we celebrated the full rollout of our multi-award-winning multimodal and generative AI indexing technology, MXT-1.
Several Newsbridge customers are early adopters of MXT-1, a patented indexing technology that understands and describes what’s inside video files. By identifying people, translating and transcribing dialogue, identifying logos, objects, landmarks, context and on-screen text, MXT-1 can describe videos shot-by-shot and produce comprehensive summaries. The AI enables organizations to index their media collections at scale and make their assets fully searchable.
One such MXT-1 early adopter, online publishing powerhouse Le Parisien, challenged us to go one step further with our technology.
What if we could take the AI’s ability to understand and describe media, and use the information generated to automatically create the SEO-optimized titles, chapters and descriptions required for online video publishing?
Le Parisien’s Video Technical and Creative Director Guillaume Otzenberger joined us in Amsterdam to demonstrate the result of summer tests of MXT-1 in his department’s video workflows.
Le Parisien is a French news media brand covering global and local news both in print—with 2.5 million daily readers— and online, where it achieves an incredible 170 million monthly video views. The company has used our AI-powered Cloud Media Hub for more than two years for archiving and publishing its video content. The release of MXT-1 was an opportunity for Le Parisien to gain further efficiencies.
However, Otzenberger’s first message to the IBC audience was “Don’t panic!” He’s not planning to replace his team of 15+ talented video journalists with machines. “We have a very clear goal when it comes to how we want to use AI, and how we won’t use it.”
Guillaume summarized Le Parisien’s ethical principles on AI as follows.
Le Parisien’s stance on AI aligns with Newsbridge’s approach to developing its technology ethically and in a socially responsible way. We believe that there will always be a need for human oversight when using AI, especially in editorial contexts.
In the fast-paced world of online news, journalists are under pressure to get their stories out as quickly as possible to maximize audience engagement. The autonomous nature of how Le Parisien’s journalists work really puts them to the test.
“Our video journalists are operating in a very modern way,” explained Otzenberger. “Most of the time it's a single person with a camera. They do the shooting, the interviews, manage the picture and sound quality. They prepare the story copy and edit the video with Adobe Premiere and several other tools.”
After all of this work, the finished video needs to be published to Le Parisien’s online platforms—including YouTube and Dailymotion—and importantly, optimized for SEO, with all relevant keyword tags, hashtags and content descriptions. This is where precious time is lost.
On average, Le Parisien publishes 20 videos per day. Each video takes at least 15 minutes —sometimes as long as 30 minutes—to manually summarize and upload. We identified this point in Le Parisien’s workflow as one that could be greatly accelerated with MXT-1.
Le Parisien’s tests with MXT-1 took place in the summer of 2023 with the objective of automatically generating video content titles, descriptions and time-coded chaptering. The SEO-optimized videos could then be published to YouTube and Dailymotion directly from Newsbridge.
Fine-tuning was conducted to ensure MXT-1’s output reflected Le Parisien’s editorial tone of voice and the optimal word count for video platforms.
A side panel added to the Media Hub interface enabled journalists to generate the desired copy for their final video edits, be it timecodes, descriptions, hashtags or headlines. Users could choose to generate one or multiple elements and immediately review and edit the text as required.
“The MXT-1 description is amazing,” said Otzenberger. “The AI understands the level of detail we need to add to a video to optimize it for search engines. The hashtags and chapters that MXT-1 generates are also very good. It’s producing all the elements that are best practice for maximum views.”
Le Parisien’s summer tests with MXT-1 also demonstrated that an incredible amount of time could be saved. Using the AI to generate titles, descriptions and video chapters and then making any necessary edits took the Le Parisien team on average three minutes. Creating these content elements manually takes at least 15 minutes per video.
When it came to the work of his journalists, Otzenberger was clear about where he wanted them to focus their time and energy. “The journalism, the story investigation, all of that hard work is done by humans. What the AI is doing is removing a repetitive, administrative task from their workflow and enabling them to publish much faster.”
Audience and story quality is key to Le Parisien’s growth strategy, and the company is constantly exploring new storytelling formats to increase audience engagement. The Le Parisien Facebook and Instagram accounts are now integrated with Newsbridge for automatic video publishing, and TikTok is next.
When it came to evaluating the usefulness of AI for his video journalists, Otzenberger concluded, “MXT-1 is a tool that really gives us an edge.”
Looking to step up your content production workflows? Contact us and ask for a demo of MXT-1.