© Nicholas Cappello / Unsplash

30 March 2022

The rbb microcosm is breeding AI projects for radio and TV.

At first glance, the evening news and kids’ cartoons couldn’t be more different. Behind the scenes, though, they’ve got more in common than many viewers might suspect: For both programmes there’s no human cutter sitting at a desk for hours on end to put different snippets together into a meaningful video. This tedious process has been taken over by an AI-based system called ‘Auto Cut’, which combines video and text mining with compositional criteria like relevance and continuity editing to produce news programmes and episodes of children’s cartoons in no time at all. Automation not only saves an enormous amount of time and delivers more output; it also gives the editorial team more space for creative processes. Our principle at the rbb is to always keep an eye on potential improvements to workflows using new technologies’, explains Susanne Büchting, head of the department of Technical Innovation Management (TIM) at the broadcasting company Rundfunk Berlin-Brandenburg (rbb), responsible for developing and implementing Auto Cut. ‘One thing is especially important to us at the rbb: AI supports people and doesn’t compete with them. AI makes our work easier and creates added value for automating and processing large quantities of data in particular’, according to Büchting, who has been working at the rbb for 18 years.


Susanne Büchting © RBB

AI: ‘A normal part of rbb employees’ day-to-day work’

AI has been a hot topic at the Berlin and Brandenburg state broadcasting company since 2019 – and not just on a theoretical level, Büchting is keen to stress. On the contrary: the 15-strong TIM team develops concrete project ideas with editorial teams and tests their relevance and their implementability. Along with Auto Cut, the rbb is working on six more future-focused projects. ‘To give you a visible and audible example, our weather and traffic updates in the rbb-Inforadio app are currently available as audio thanks to automated text-to-speech software and a synthetic voice that we had specially produced’. The project has been running for two years already now. Like with Auto Cut, though, most of this goes on behind the scenes. ‘AI is a normal part of rbb employees’ day-to-day work now, including on the production side, like producing image and video material’, the TIM boss explains. Infographics are generated automatically to quickly visualise COVID cases, unemployment statistics and football leagues for a wide range of platforms. Meanwhile, auto-reframing creates video clips that are adapted to meet the specific requirements of platforms like Facebook, Instagram and individual websites. An AI-based analytical process deals with centring objects for this. The rbb is also working on constructing a material database of personalities and landmarks in the region: Automated video analysis and certain metadata help the AI recognise people, objects, and geographic and demographic parameters. The aim is to make finding video material quicker and more efficient in the future: ‘Obviously at TIM we’re in touch with companies and we test out their AI services to see if we can include them in our production work and optimise both processes and results’, Büchting adds.

The special role of public broadcasting

The rbb isn’t the only company with a keen interest in AI: Chatbots, SEO, automated weather, financial and traffic updates, and graphics assessing the course of play in sports events and live broadcasts – artificial intelligence is well-established in the media industry, as XPLR: MEDIA in Bavaria’s AI report proves. ‘The subjects we work with are similar to ones other media companies deal with too’, Susanne Büchting explains. She does emphasise one difference, though: ‘Up to now, private broadcasters have used AI to place advertising in line with users’ interests – that’s not what we’re aiming for!’ The use of AI is free from political and commercial interests at the rbb. ‘We use AI to fulfil our responsibility, that is, in the best interests of the community’. In her eyes, that is the role of a public broadcaster. ‘It takes a lot of transparency and high quality standards in data protection, etc.’. This particular responsibility is made clear in the European research project ResKriVer, which started in 2021 and will run until May 2024: The rbb intends to work together with research and industry experts to develop a communication and information platform for robust, crisis-relevant supply networks. Büchting describes the project like this: ‘The idea is that artificial intelligence and data technology will help meet the specific communication needs posed by crises and catastrophes’. The systems are intended to make verified data on crises and catastrophes available and to support crisis-relevant data processing. An interactive, personalised and tailor-made selection of information is to be developed, based on the platform’s wide-ranging data and intelligently linked to relevant actors. ‘Today’s crises are making clearer than ever just how important it is to inform the population comprehensively and objectively about the particular situation, to sequence and evaluate events and to offer the population support they can trust’, in the expert’s opinion. ‘We hope that using AI will allow us to take advantage of optimised social-media listening in the field of crisis communication’.

ReTV, a pioneering project

ResKriVer isn’t the first time the rbb has got involved in an international research project on the topic of AI. ‘There are AI components in almost every project, and that’s been the case for several years now’, Susanne Büchting, who has led TIM for the past seven years, confirms. Back in 2018 the rbb plus five partners – companies, institutions and an archive – from five different states took part in the ReTV project as part of the EU’s ‘Horizon 2020’ programme. The aim was to provide users with tailored video content. ‘For this reason we really focused on researching users’ interests and behaviour patterns’, says the TIM boss. ‘Using machine learning and AI helped us develop a prototype system to automatically recognise video content and produce video summaries or new videos. This experience was unbelievably useful in the projects that followed, like our internal project Auto Cut, which now deals with automatically cutting videos’. Experiences and expertise picked up in the project are coming up right on cue in ResKriVer now. ‘We’re now also applying for new projects in which AI will continue to play a central role’, Büchting adds.

There’s certainly no risk of getting bored at TIM, where ongoing projects are constantly evolving and expanding. AI-supported evaluation of additional programme and user data, for example, is to help provide new insights into target groups and viewers’ interests in order to optimise TV scheduling. ‘This showed us that we first had to develop our own AI’, Susanne Büchting says. ‘There were lots of key factors that we had to link together to start with, and we identified more than twenty data sources that are important for scheduling. The project is going to keep us busy for a few more months yet’. Algorithms, too, are expected to support professional journalism even more in future: Detecting topics, helping with research and putting content together according to specific parameters. With all this digitisation, though, there’s one thing she’s keen to clarify: ‘We don’t use AI for the sake of it, we use it to do our job better and in a more modern way. If in doubt we can use AI to work more, work more quickly or work more accurately, and to improve our part in shaping opinions and improving society as a whole.’