A round table of the Collettivo Chiaroscuro together with HAI Human & Artificial Imagination (AI production service powered by EDI), to explore what changes in the work of the director of photography with the arrival of artificial intelligence.
Curated by Alessandro Chiodo and Stefano De Pieri
Warm thanks to ANICA for hosting the event in its cinema screening room.
When we decided to organise this round table on AI and cinematography, the idea was not to hold yet another meeting on artificial intelligence. There are plenty of those already. Too many, perhaps. Often very generic, often split between blind enthusiasm and outright panic. We were interested in something else: understanding what AI means for our craft. For those who work with light, with the set, with lenses, with the physical substance of the image. For those who, even today, still believe that cinematography means getting your hands dirty. That is why we wanted a small gathering — closer to a working session than a panel. A room not designed to turn the subject into a spectacle, but to get into the substance of it. With practical questions, even uncomfortable ones: what can AI do for a DoP today? Where does it enter the workflow? How much does it cost? Who controls it? What risks does it open up in terms of authorship and legal liability? And above all: how do we prevent it from becoming a tool managed by others, over our heads?
With us were Stefano Leoni — partner at EDI and co-founder, together with Francesco Grisi and Francesco Pepe, of HAI Human Artificial Imagination, EDI’s AI production service — Fabio Scottellaro, Head of Lab at EDI; Elisabetta Rocca, Head of Production; and lawyer Donatella Mugnano of Studio Legale Miccichè e Patriarca, to address what is perhaps the most pressing issue of all: rights, contracts, protection of the work, data use, and tax credit.
The starting point was straightforward: AI is not an intelligence that decides in our place. It is a tool. But a very powerful one, and like all powerful tools it changes the way an image is prepared, produced and defended. Stefano Leoni put it clearly: AI can shorten decision-making time in pre-production, especially when cinematography and VFX need to engage with complex scenes, difficult locations, limited budgets or aesthetic choices that would normally come too late. Instead of arriving on set and saying “we’ll fix it in post”, we can begin analysing the problems earlier. We can manipulate, test, simulate and put a visual idea through its paces before shooting it.
This does not mean replacing the set. It means arriving on set with a clearer understanding of what is actually needed. And here, in our view, the first real question arises: we need to begin building a new grammar with production. When we prepare a film or a series, it is no longer enough to ask only how many days we have, which camera we will use, which locations, how much electrical equipment. We need to start asking: what can we develop beforehand with artificial intelligence? Where can it help us visualise? Where can it free up resources? And where, on the other hand, does it risk creating the wrong expectations? Because the risk exists. We already know this. Many have experienced presenting to a client or a producer AI-generated images, moodboards, animatics or previsualisations that are more beautiful, more seductive, more “finished” than what the actual budget allows you to shoot. At that point AI no longer helps: it becomes a weapon against the set. The answer, however, cannot be not to use it. The answer is to use it better.
The thought that stayed with us most was this: we need to bring artificial intelligence to help us achieve an analogue result. Not a perfect digital image, polished, self-sufficient, closed in on itself. But an image that still has imperfection, body, friction, substance. That quality which CGI has always had to recapture: dirtying the calculation to make it feel real. In this sense AI does not eliminate the DoP’s work — it can make it even more necessary. Because someone will still need to know what is missing from that image. Where the light falls. Where it must not fall. How credible a background needs to be. How imperfect a backlight must be. How much weight a lens choice carries, a grain, a density, a colour.
The production dimension is equally central. AI can make possible productions that previously never got off the ground — not because it is free, and this must be stated clearly, but because it can allow more to be done with contained budgets. A mid-range film, a series, an Italian or European project that cannot afford a certain kind of construction, set extension, science fiction, or environmental transformation, can find new paths. But this possibility needs to be governed. If it becomes simply a way to cut crew, shooting days, departments and expertise, then it is not innovation: it is impoverishment disguised as the future. If instead it allows the core of the work to be protected — the actors, the set, the direction, the cinematography — and shifts some of the production complexity elsewhere, then it can become a genuinely positive tool.
From here we arrived at another question that interests us greatly as a Collettivo: is a new figure needed on set? During our internal discussions we tried to imagine a bridge between the camera department and the VFX department. A figure we have called — perhaps provisionally, but very clearly — AIT: Artificial Image Technician. Something that stands in relation to AI as the DIT has been, and continues to be, for the digital image workflow. Not a figure parachuted in from outside to generate “beautiful things” at random. But someone inside the cinematographic process, capable of engaging with the DoP, DIT, VFX Supervisor, laboratory, colourist and production. Someone who knows how to control image consistency, work on proxies, previsualisations, backgrounds, virtual production, on what arrives in camera and on what will arrive in post. Stefano Leoni confirmed that such a figure makes sense, especially where AI is beginning to enter the actual set. The first terrain will probably be virtual production: generated or reworked backgrounds, modified environments, images that must respond to real light and not contradict it. Because the central point remains always the same: light must guide the process, not be subjected to it.
The technical limitations still exist today. Resolution, native 4K, detail, HDR, consistency between shot and generated footage, the need for compositing. No tool today truly replaces a complete cinematographic workflow. The generated image often already looks post-produced, already “cinematic” in a generic way — and precisely for this reason it needs to be walked back, de-artificialised, made less polished and more credible. But these limitations are changing fast. Six months, in this field, is already an era. That is why we cannot afford to wait for the process to be defined elsewhere.
The legal dimension opened a second front. Law 132/2025, we were told, reinforces a fundamental principle: works created with the assistance of artificial intelligence can only be protected if a recognisable human creative contribution remains. This means that human work must be documented. Prompts, revisions, software used, creative steps, reworkings: everything that demonstrates that the image has not simply come out of a machine, but is the result of a directorial decision. The same applies to datasets, to the use of images, faces, voices, likenesses, and protected materials. We can no longer afford to be careless. Free licences are often not truly free: they grant rights, use the content uploaded, offer no guarantee of confidentiality, and are not suited to professional use. If we are working on films, series, or production materials, we must demand tools with professional, enterprise-grade, verifiable licences. This introduces a new concept for creatives and beyond: the ownership of the Digital Data Asset — a unique asset created specifically for a given project.
And then there is the most delicate issue for us: the protection of the director of photography’s contribution. Today a system can be trained on a style, on a sensibility, on a visual recurrence. It can learn a way of lighting, of contrasting, of grading, of composing. It may do so poorly, it may do so in a homogenised way, but technically it can try. And if the process counts for less than the final result, then we need to start protecting the process. Our visual culture, our experience, our way of arriving at an image. We are not co-authors of the film under current Italian law. We know this. But we are bearers of a recognisable creative contribution. And at a time when that contribution can be absorbed, replicated, trained and commercialised, the profession must confront the problem. Contractually, culturally, politically.
By the end of the day, it seemed clear to us that the question is no longer: will AI enter cinema? It already has. The real questions are: with what grammar? With what rules? With what figures? With what responsibility? And above all: who will be authorised to make decisions about the image? For us, the answer cannot be left solely to software, producers, platforms or technical departments. It must involve those who conceive the image, prepare it, take risks with it and sign for it every day on set. Artificial intelligence can be a shortcut to doing things worse. Or it can become an opportunity to do things better — with greater awareness, greater control, greater dialogue between departments. It depends on how capable we are of engaging with this new tool, without fear and without naivety.
The Collettivo Chiaroscuro exists, among other things, for precisely this: to create spaces for dialogue where the craft does not simply undergo change, but tries to name it before it becomes a rule written by others. After the summer there will be a new stage in our exchange and our exploration of the future of cinematography.
