"Nous séquençons l’ADN de chaque film"
Dossier industrie: Tendance du marché
Marco Moretti et Antonio Emmanouilidis • Co-créateurs de Cabirya
Nous avons rencontré les concepteurs de Cabirya, première API de recherche dédiée au cinéma européen et premier moteur de recherche pour les images animées, pour discuter du projet
Cet article est disponible en anglais.
Cineuropa jumped at the opportunity to speak with Marco Moretti and Antonio Emmanouilidis, who founded Cabirya, the first search API for European movies and the first search engine for moving images. We sought to uncover a few more details on the project and to learn how the European film industry stands to benefit from its creation.
Cineuropa: What is Cabirya?
Marco Moretti: Cabiria was the title of the first blockbuster shot in Europe, and it’s also the most famous film from the golden age of Italian cinema, so this was the name we chose for the first search API for European movies and the first moving images search engine. Recent market analyses revealed that, during the pandemic, the number of subscribers to European VOD platforms exceeded 140 million. But the percentage of European audiovisual content in the nigh-on 800 VOD service catalogues available in Europe still didn’t go beyond 20%. Furthermore, the visibility given to these European productions on the various platforms’ homepages averages 18% against 58% for their American counterparts.
Could it be the case that we have a significant pool of viewers who are unable to find the huge, mind-blowing European production they’ve been looking for? This is the question we asked ourselves. We realised that there might be a need for some sort of meeting point between viewers and audiovisual products, a system which would allow them to find specific content, European in this instance.
As part of our project, we analysed around 200 of the VOD platforms available in Europe. Ultimately, the only way users can find content is via a rudimental search form, where you insert the name of the director or the film, and, sometimes, the names of the actors, the particular country of origin and the genre. Cabirya was subsequently born as a search platform for moving images. It’s intended as a very lightweight platform, a stand-alone, multi-platform search engine, or as a system integrated within the existing VOD platforms. In essence, it will generate better recommendations and will help audiences to discover European films.
How does it work?
MM: Each film is segmented into narratively significant scenes by way of cinematic-based boundaries detection technology. Each segment is analysed, from individual-shot-level to the entire scene, and is then tagged via a semi-automatic system. And what does this give us? It indicates the content in terms of people, objects, brands, texts, actions, locations, colour cast, framing, camera movements, sounds and emotions. We sequence the DNA of each film. To give you an example, a feature film generates around 1,800 tags. These tags become new gateways to the film. The system also correlates similar elements from similar scenes hailing from films from different cultures, languages and times. In this sense, the discovery of European film will acquire something of a “granular” appeal.” And who will this benefit? It will benefit film lovers, VOD platforms, distributors, producers, filmmakers, film institutes and archive centres, because it’s a way of revitalising and helping increase the circulation of catalogues which are often unexploited, chaotic or quite simply inaccessible.
Antonio Emmanouilidis: It’s no longer the public or a sizeable market that’s missing, like it used to be - there’s now half a billion of us! […] It’s not even a case of too few high-quality stories. But there is a problem in terms of language and cultural differences. It’s a crucial entry barrier. But what if we lower this barrier? For example, I like action films. There’s no doubt a French film with a brilliant action scene in it. That particular scene might be to my liking, but I’m not interested in the whole movie. Nevertheless, it will be difficult for me to access that scene without watching the entire film. The same goes for the Divine Comedy: it’s a three-volume work; some parts will interest you, others less so. But what if I gave you a tool which could point you towards the parts you’re interested in?
And then I’d be free to choose whether to take things further or not.
AE: Exactly. We believe this could also be a way of helping the public to discover new types of films.
What are the greatest technical difficulties you’ve been faced with while developing Cabirya?
AE: The first problem is tagging each film; it requires a huge amount of work. That’s why it’s never been done before. The other problem is that there’s still no way of doing it automatedly, you have to do it manually or semi-automatically, as we’re doing.
MM: Generally speaking, image and frame recognition is automatable. Action recognition, less so. What isn’t automatable, for example, is identifying the beginning of a narratively significant sequence: the cataloguer has to manually validate it.
Is your company developing the tool or are you enlisting the help of external partners?
AE: We’re taking care of the research and testing phases, but we’re working with an English start-up for the technical aspects.
Which countries will be the first to benefit from the project?
MM: The project is global by nature, but in this initial phase, which is aimed at validating the system and producing a prototype, we’re offering to incorporate the catalogues of archive centres and film libraries in various countries into Cabirya’s advanced mapping system. It’s a way of making this vital film material available outside of academic circles.
(Traduit de l'italien)
Vous avez aimé cet article ? Abonnez-vous à notre newsletter et recevez plus d'articles comme celui-ci, directement dans votre boîte mail.