Taking the Deep Dive: how an AI commentator can improve viewers’ experience
Here we will tell you how we created Cinephile — an AI commentator for films and TV series. Cinephile was created by Watchers and can be integrated into any video platform.
What do you want to know while watching films? In order to answer most of the questions that might occur to you while watching films or TV shows you need Google, additional materials or a friend who is an insider in the film industry. From time to time, we all rewatch films to focus on costume design, look more closely for the boundary between science and fiction. We download Shazam to figure out what song a character was listening to while driving…and, of course, we never quite manage to open that App in time!
There can be a plethora of questions about the films we watch, and all of them have clear answers, which viewers have to do a lot of work to get, and may have neither the time nor the skills — let alone the resources — to answer all of their own questions. So it occurred to us: “What if we were to do the job for them?”
X Ray: remember who that actor was and buy his watch
In 2012, the Amazon Prime Video VoD service announced the launch of a feature which helps viewers ID an actor on screen and see a list of the other films in which they played without exiting the viewing experience — just by pressing pause. At last! This solution was called X Ray. Users who had read books on Amazon Kindle were already familiar with X Ray — three years earlier, Amazon books had done the same thing with its service — the result of a collaboration between Kindle and Wikipedia.
In the case of Prime Video, the feature was based on IMDb, which had already been owned by Amazon since 1998. The X Ray feature would simply take information about the broadcast from the popular IMDb database using face recognition technology.
That’s what you see when you pause while watching Amazon Prime with X-Ray (Image by Business Insider)
The new feature attracted both film nerds and ordinary viewers, who simply didn’t have any desire to look for additional information and were now able to get it with no effort. Over time, X Ray began to offer information from other sections of IMDb, such as the Trivia section. Later, X Ray started to recognize songs and lead viewers to Amazon Music. In some cases, it even identifies clothes and equipment on actors and recommends them for purchase via Amazon marketplace! In summary, X Ray has morphed from an extra source of information on cast into a service with smart-market functions, actively promoting Amazon services and developing the whole company’s “ecosystem”.
Some local services have tried to reproduce X Ray’s success and deep dive functions, but it hasn’t happened so well. To launch such features even for one film, or develop sophisticated AI technologies, an extremely large amount of work is required.
Why not dive deeper?
It goes without saying that the opportunity to be prompted about who is who on screen is amazing. Even when we’d just come up with the Watchers concept, we realized that social watching should include extended information capabilities. And if we want AI capabilities, these should be capable of much more than info-borrowing from open-source materials.
Watchers is a white label solution, which provides the ability to create online rooms on your video-platform. In these online rooms users will be able to watch content simultaneously while chatting by text or voice. At first, Watchers was created as a b2c app and now it is a SaaS being used by companies to show-case their content and to boost their business making use of social mechanisms.
Otherwise, the difference between watching a film with the deep dive features and watching while reading Wikipedia in parallel is going to be minimal. Moreover, using deep dive services viewers need not press ‘pause’ while garnering the information. With Wiki, users can simply read during organic pauses. Thus we decided that commentaries are best when linked to сertain moments within a film, rather than shown on request. Afterall, viewers may also simply not know that we have something of value to show them.
In addition to the problem with pauses, there is another one, which is text style. Users are accustomed to the style used by their favourite authors, bloggers and even services — because all of these “speak in their language”. Wikipedia and open-source databases make no attempt to be approachable to the user. Interrupting viewers’ experience with dry language that doesn’t suit them just cannot be a good idea, no matter how good the intention.
It has also become very clear that cross-promotion — whether done inside one large service provider (like Amazon) or when implemented as an inter-service solution, can be developed much more thoroughly. After all, in addition to highlighting product placements which already exist within a film, you can also help viewers satisfy newly-introduced appetites brought about by a film or TV series, itself.
All that said, we launched our own AI commentator to truly create a deeper watching experience. It’s not that we want to replace Wikipedia and IMDb; it’s much more that we want to give users absolutely unique content — intentionally linked with specific titles and created by experts both from the world of cinema and across various other fields — relating to particular moments in the viewing experience.
Cinephile is a new, more attentive and personal type of immersion experience.
What viewers are looking for
It’s clear that not every film or TV show is a perfect fit with this type of deep dive feature. There is very little to discuss in some films. In some cases, frequent messages would only break viewers’ suspense or even ruin the pleasure of watching! Our first task was to learn what sort of content is ideal for commentary.
And viewers, themselves, are just as diverse as film genres. Some people prepare before viewing content, while others prefer to know nothing in advance. Every person comes with their own background and has their own viewpoint. Even attitudes to suspense can be different. Some viewers like to look away while watching scary films because this helps them be less frightened. Others specifically want to experience a feeling of horror, believing that this is the entire point of watching such a film.
We analyzed various platforms where viewers shared their impressions of films and the whole watching experience. We studied users’ reviews, social media, recalls on VoD-services, and — of course — Reddit, which is the place to discuss everything. Our next step was to ask the viewers themselves what types of films they prefer to watch with additional content.
Our test group included both men and women between the ages of 18 and 38, who use paid online services to watch video content, who in one way or another have some knowledge of film genres and directors, and who regularly watch video content for entertainment purposes.
When the viewers’ actual demand for additional content began to take shape, we categorized a number of films to work with, keeping in mind that sometimes our categories would intersect.
● Historical films and biopics
● Films with open endings or confusing plots
● Stories with a broad cultural context
● Films made from comics
● Stories which have other versions (adaptations, remakes)
● Sequels and prequels, which are watchable as a single series
The types of films which proved inappropriate for AI commentary, according to our respondents, were action movies with dynamic plots, thrillers, and horror films.
Users agreed that watching patterns can differ by film, meaning that the additional information offered, as well as its frequency and general attitude, should be carefully adjusted to the specific watching experience. Different respondents wanted content of different complexity levels. Our more expert users required deeper analysis of films, while our less expert fans wanted simpler and entertaining factoids.
Say something “in cinephilic”
From the get-go, our AI commentator was given the named Cinephile, just like the most immersed of all immersed viewers. We want Cinephile to be as specific and personalized as possible, meaning it should be able to adjust to any film it expands upon, and to each and every specific viewer who switches it on.
The term ‘cinephile’ is derived from the French word cinema and the Greek word philia, and means “film-lover”. The term cinephile was first connected with film fans in 1940s, post-war France, at a time when foreign cinema had just come to that country and local films began to be aired more often. Young people began to watch films with great excitement. These cinephiles didn’t only watch films — they also discussed and wrote about them. In 1951, the members of different cinephilic groups in Paris began to publish the most influential magazine of that time about cinema — “Cahiers du Cinema”. Many of these young people became film critics, and later, even directors. When the wave of cinephilia began to ebb in France, the phenomenon nevertheless began to spread around the globe. Nowadays, people who call themselves cinephiles don’t just love films, they are interested in film criticism and film studies.
Cinephile content for each film is created by experts in cinema and other fields of interest which crisscross the plot. For example, the Cinephile material for the criminal drama about serial killer Ted Bundy, entitled “Extremely Wicked, Shockingly Evil and Vile”, was not only about filmmaking, but addressed how the real Ted Bundy looked and acted, as well as the historical actions of the police and court. Fashion journalists provided commentary about fashion, telling viewers more about the brands worn by characters and what directors want to say with their particular choices.
Sounds a bit formidable? But after all, viewers want to watch a film and not read a report. Cinephile messages sent to users while watching must therefore be short and clear, which is why our editorial team converts the texts provided by experts into material that is easy to read, and checks carefully whether they are appropriate for transmission via AI commentator, making any necessary changes. Editors correct everything from message content to timing.
In summary, Cinephile is not only a friendly film critic, but a very able teaching, marketing and engaging business instrument.
Alexey Filippov, editor of Cinema Art magazine, Monday Carma podcast host
I think that the modern avalanche of content, emotions and information is the reason why curated selection of facts and emphases is being so demanded.
Of course, everybody has their own relationship with home viewing, sometimes this process looks like a reading of an interactive book. We get distracted to find an interesting fact, name of the actor or even to discover the city map, dish recipes, genre history etc. Thus viewers who don’t like it won’t use such services.
For most viewers watching a film is not a daily routine or the inner work. It is a quest for emotional experience, for some kind of event. Amusing interpretation of an expert can turn out to be a rare event and open new sides of a well-known film.
A viewer forms their own interpretation without anyone’s help. It is easy to find basic factoids about the film without additional guidance, a bit of curiosity will do the job. After homebrew research it is more valuable to return to a known film, to watch it from a new point of view. Serendipitous facts and remarks can evolve perception faster. It is a part of the magic of cinema, isn’t it? Or the magic of culture.
Thus, once again, more films and more experts means more opportunities. Use them!
If you want to learn more about Cinephile AI commentator or to integrate it into your platform, сontact us: https://watchers.io
Text by Alina Kuzio