How AI can empower broadcasters in the changing media landscape

A deep dive on the four different areas in which AI can be benefitial for broadcasters

Broadcasters are slowly adopting AI technology in their day-to-day operations. That emerged from this year’s EBU Production Technology Seminar (EBU PTS 2019) in Geneva, which Media Distillery also attended. Several broadcasters presented their experiences with AI, indicating that they are becoming aware of the opportunities AI can offer. However, from our own talks with broadcasters, it also became clear that those in the industry are not always aware of what added value AI can bring to broadcasts. In this post, we therefore highlight four different areas in which AI can benefit broadcasters: producing content, improving content accessibility, enabling advanced advertising opportunities and improving viewing experiences.

Content Archives: know exactly what’s inside

Many broadcasters own large content archives, containing programs from decades of broadcasting. These archives are often a black box because proper metadata is missing. That makes it difficult to reuse archived content, for instance in the case of reusing footage for a news story or making a compilation of a celebrity or athlete for an entertainment program. At EBU for example, Swiss broadcaster RTS expressed having little or no metadata for 70% of its content. The lack of data stems from the fact that creating it for content has primarily been a manual, and therefore time and resource-intensive task.

However, with AI-based content analysis tools, a lot of metadata can nowadays be generated automatically. Speech-to-text tools allow audio dialogue to be indexed as text, which then can be used for video search, but also for creating synopses or the key topics of a video. Face and object recognition tools make it possible to detect who or what’s inside a video. Through automatic content analysis, broadcasters get a detailed view of what’s inside their content, transforming the archive from a black box into a usable asset for content production.

Improving content accessibility

Content accessibility is an important requirement for broadcasters. It’s not only important for them to serve viewers with special needs, but more recently also important that content can be viewed while audio playback is muted, for instance when people view content while commuting. To ensure that quality meets broadcast requirements, accessibility measures such as subtitles, closed captions and sign language are primarily created manually. Automated captioning systems already exist for some time, and can for instance be seen on YouTube, but their quality does not always meet broadcast standards. In addition, captions or subtitles summarize dialogue, rather than being a verbatim copy of spoken language, which tools still struggle to achieve. New AI-based automatic speech-recognition solutions offer a higher level of accuracy, making it possible to create subtitles by respeaking for TV, or for online playback. This means that with less manual effort more content can be provided with captions. As we see new developments in AI, we’ll also see other forms of content accessibility improve and require less human intervention, to the benefit of all consumers.

Enabling advanced advertising

Many broadcasters depend on advertising revenue, so they need to ensure they keep offering value to advertisers. For a long time, advertising opportunities for TV lacked something that has for many years been the standard for the online world – the ability to target ads to a specific consumer segment. Developments such as DVB Targeted Advertising, enhancements to HbbTV as well as regional initiatives such as those in the Netherlands are changing this. By having a deep understanding of broadcast content, more advanced advertising scenarios will also become possible in the future. Here are a few of the areas we’re working on at Media Distillery:

Contextual advertising: by analysing the video content, we can detect ad blocks or ads, so that they can be replaced by more relevant ads for the context. Better ad placement: mid-roll advertisements impact the user experience, especially when their placement is poor, such as in the middle of a scene, or worse – in the middle of a sentence (yes, that happens frequently). By automatic content segmentation, the optimal placement of an ad can be determined, with reduces user annoyance and improves advertiser satisfaction. Ad-detection to enable ad-replacement: some broadcast content lacks metadata about the exact start or stop of an advertisement block or individual ad. AI can help detect the exact start and stop times.

Improving viewing experiences

The days when consumers could only view content via broadcast are a thing of the past. Broadband internet access has made it possible to deliver video content Over-the-Top with a video quality similar to or even better than broadcast. The transition to OTT delivery is beneficial not just for consumers, but also for broadcasters; it allows them to serve and build a relationship with consumers directly. However, it has also opened broadcasters up to competition from new OTT players, that have raised the UX bar for video services. “Digital native” consumers are expecting a different experience when accessing broadcast TV content, and AI can help achieve this. When you know what’s inside the content, many UX improvements become possible making your video service more enjoyable:

You can offer advanced search features, where viewers can finally find something inside video and don’t get the dreaded “0 results found” just because there’s insufficient metadata. Combining advanced search features with voice control interfaces will result in a new, natural way to enjoy content. You can offer better content discovery mechanisms. For instance, by offering tailored video thumbnails/still images that actually help the viewer select the right content. You can enable binge viewing: let people skip the intros and outros of programs, in order to avoid viewer annoyance at having to do that using the remote control. Some OTT players already have solutions in this area.

Broadcasters already have the quality of content that attracts viewers, but now need to focus on improving user experience to avoid consumers turning their attention to other parties.

These are just some examples of how AI can benefit broadcasters today or in the long-term. Or to put it slightly differently – these developments are already in play, so as a broadcaster you should consider embracing the opportunities of AI if you are not already doing so. And this is just the tip of the iceberg since we haven’t even touched upon camera tracking, creating trailers, preproduction and video encoding.

April 18, 2019

Published by:

Martin Prins

Head of Product