Natural Language Processing
Intelligent Search
Managing large volumes of TV content can make it difficult and time-consuming to organise programmes into clear, searchable chapters. For VRT, the challenge was to make its video content easier to navigate and more accessible, without relying on manual chaptering for every programme.
VRT MAX, the online video platform of the Flemish public broadcaster, offers viewers the ability to rewatch their shows and even jump to specific sections. A great feature for their customers, but an intensive manual process behind the scenes. It demanded a lot of time and effort from the already busy editors, so VRT wanted to find a way to make chaptering easier and more efficient.
1
Increased efficiency allowing editors to focus on higher-value tasks.
2
Optimised workflow leading to time and cost savings.
3
Enhanced metadata provides insights and opens new possibilities.
To develop the automatic chaptering tool, Brainjar experimented with a range of machine learning techniques, refining their approach in several iterations. They came up with several algorithms that are capable of automatically dividing episodes of shows like "De Afspraak" into distinct chapters. Besides chaptering, Brainjar also enriched VRT's videos with valuable metadata that unlocks new possibilities for future feature development.
Right from the start, knowledge transfer was a top priority for VRT. They wanted to be actively involved in the process, since they would be able to continue working with the tool independently. The experts at Brainjar collaborated closely with VRT's internal team, sharing knowledge and ensuring a smooth handover that allows them to further develop the tool into a product and apply it across a wider range of content.
While the end-user might not see a difference at first glance, this AI-powered solution is a game-changer behind the scenes, saving VRT's editors valuable time and allowing them to focus on more creative and engaging aspects of their work.
.jpg)
%20copy.webp)
