One consequence of the swing to online video consumption is that the technology of delivery and display can no longer be divorced from content or its creation. It has hastened the convergence between entertainment and communications and the winners are those who have milked this trend most effectively. Netflix is the perfect example, having become the global force it is now on the back of original content production geared to online audiences able to watch when and where they like. Ironically this has had consequences many traditionalists or even luddites would welcome, stimulating consumption of historical dramas, documentaries and fiction.
The latest confirmation of this trend has come from French audience measurement firm Mediametre, reporting that fiction has benefited more than any other genre from the growth of catch up and multiscreen viewing. Such content trends have also helped integrate viewing across all devices including the main TV, the agency considers. This conclusion is reasonable and again is how the big SVoD players have thrived, by stimulating binge viewing and gaining eyeball time across all screens.
At the same time though there is a distinction between screens, as well as by location and time of day, in content choices people make, which must be factored in to recommendation and also ad targeting. Such choices are not always obvious, for while there is a tendency to watch shorter form material on smaller screens there is also a lot of catch up viewing of feature length films, or resuming something started the previous night on the big screen.
So, while Facebook and others have commissioned short movies and snackable content for mobile devices, they are also investing in longer form content. Decisions over what content to create are themselves increasingly data driven, which again is why the big tech giants have gained so much ground and why traditional content producers, including some Hollywood studios, are having to respond by going direct to the consumer. That led Disney to pull its content from Netflix in preparation for launching its own OTT service. It is also pushing traditional public service broadcasters to authenticate users of their online portals so that they can start drawing in data about preferences and individual viewing behavior.
The next stage of convergence may be even more profound as technology increasingly infiltrates the way content is made as well as how it is promoted, distributed and consumed. This is where AI and AR (Augmented Reality) are coming together to help create content of all genres, not just movies but even in the longer-term live sports, as well as documentaries. So far, the impact has been confined to movie trailers, where IBM’s Watson got in first in 2016 with the first entirely AI-produced trailer for the film Morgan. Of course, a trailer is derived from the movie already made and so in this case no original content was created by the AI algorithms. But since creativity involves reassembling and expanding on past works, in the generation of trailers the potential of AI to go further and start producing original works has already been demonstrated and will happen. No doubt the first such movies will not be very good but over time AI will come to play a major role in content creation.
Coming back to the trailer, machine learning is also being employed to predict from that how successful the associated movie will be. This has been done by 20th Century Fox in partnership with Google for its experimental “movie attendance prediction and recommendation system”, codenamed Merlin, using the latter’s servers and open-source AI framework TensorFlow. It has also used Google’s YouTube-8M, a publicly available dataset of YouTube videos, for training. This dataset includes a pre-trained model from Google able to identify video features like color, illumination, many types of faces, thousands of objects, and several landscape categories. Merlin then extracts these predefined characteristics as the first step in determining which elements of the trailer correlate most closely with viewers’ preferences, so that it can then predict audiences.
There is then potential for this to feed back firstly into the trailer creation in order to define and maximize audiences. Given the trailer is constrained by the movie content itself, the next step could be to influence the underlying story or at least steer its production, or choice of actors for example.
The role of AI will not be confined to so called creative content but will extend to live sports and documentaries in different contexts, where AR will also figure. For live sports there has already been demonstration of AR to produce clips, almost like action replays, of what might have happened if say a previous event such as a foul had not happened. Whether such “what if” replays catch on remains to be seen but they will surely become popular with studio pundits aiming to make points not backed up by the actual action that took place.
In documentaries there is again scope for applying CGI to explore alternative scenarios or versions of history, interwoven perhaps with real footage. The direction this takes will in turn be highly data driven.