The 2025 edition conference confirmed that the industry is moving past speculative fervor and is now engaged in the difficult, practical work of building new pipelines, managing new team dynamics, and defining the creative-technical roles of the future.

The conference was once again curated and moderated by Sten-Kristian Saluveer, CEO of Storytek / Innovationlab and Head of Cannes Next at Marche Du Film - Festival de Cannes.

For the opening keynote speaker of Picture This 2025, Paul Trillo, pioneering filmmaker and a globally acclaimed filmmaking star, took the stage to challenge prevailing assumptions about artificial intelligence in cinema. As a multi-disciplinary artist, writer, and director celebrated for his decade-long work with AI—including the acclaimed Sora project—Trillo’s opening keynote address was a bullseye, as it situated itself in the opportunities AI brings to the industry yet at the same time was clear of its limitations and the needt o ground artistic integrity.

From the outset, Trillo set a provocative tone, declaring, “Most people are doing AI wrong… the technology is still being fundamentally underutilized.” He argued that the current industry obsession with using AI to mimic cinematic reality has led to a creative dead end and massive slop. Instead of pushing boundaries, he observed, “57% of the media on the internet is AI generated… it’s a remix of something else. It’s honestly getting really boring and not really pushing culture forward.” This, he warned, risks a “homogenization of aesthetics, homogenization of thought, of ideas,” where algorithms reward only the most easily digested content.

Trillo’s critique extended to the culture of convenience that AI has fostered. “Craft is being devalued so that we just pump out content endlessly,” he noted, referencing the rise of “fast food for the eyes.” He illustrated this with examples of mass-produced AI videos, where the speed of creation is prized above all else: “It doesn’t matter how much craft or thought was put into it. It just matters that it was fast.”

Trillo’s own journey as an artist has been defined by a willingness to “misuse” technology. He recounted early experiments with chroma key effects and machine learning models, describing how “using the tools in the "wrong way" led to all sorts of different experiments.” For Trillo, every new technology represents “a new visual opportunity… if technology keeps advancing, we can actually discover visuals and ideas and concepts that haven’t been done before.”

The second session of the conference focused on the practical applications of AI in reshaping the historical role of virtual production through AI for Scandinavian productions. Pioneering VP and VFX supervisor Martin Madsen delivered an engaging keynote on the evolving role of artificial intelligence in visual effects (VFX) for Scandinavian independent film. Drawing on nearly two decades of experience across international film and television, Martin positioned AI not as a tool, but as a creative collaborator—a “synthetic brain with the potential of a child’s, one that can be shaped and guided by human intent.”

In his keynote, he provided both strategic context and practical examples. Martin began by reframing the industry’s relationship with AI. He argued that, “AI should not be used to synthesize art or create stories entirely. It needs to be a toolset—something we can use to strengthen our vision, to create our mission more cheaply, better, and maybe also for inspiration.”

He likened AI to a four-year-old child: “It knows a lot about the world, but there’s also a lot it doesn’t know. That’s our job—to train it into creating exactly the visuals we want.”

A central theme of his session was the importance of context and control. Martin had shot a variety of footage to illustrate how AI’s power in VFX lies in its ability to understand and manipulate context, enabling rapid iteration and creative exploration. “Context is what makes this all work,” he noted, describing how AI can infer details about a scene—such as the symmetry of a house or the way overgrown plants might appear after a century. “We can iterate so much faster… you can push out many different versions and find out where you want to end up.”

Ethics and the Use of IP in AI

The discussion began with a question about the ethics of using copyrighted material in AI-generated content. Paul Trillo addressed the issue directly:

“Fan films have existed for many, many years... I don't want to discourage that. I take issue with, specifically, how OpenAI has used it as a marketing tool... They had a conversation, they're like, ‘Oh, we created Ghibli quite well.’ And you saw OpenAI's valuation shoot up after that. So they are using copyrighted IP as a way of engaging things for their own benefit and raising more money off of it.”

AI in Business and Creative Workflows

An audience member asked how the speakers use AI for business and administration. Paul Trillo shared:

“We've used it a lot to kick out, like, to brainstorm press releases... If I can go back and forth with ChatGPT to get the best press release and then hand that over to the PR firm, then the story gets far better communicated.”

Creative Decision-Making and AI

Moderator Sten Saluveer asked how the panelists decide when to use AI. Paul Trillo reflected:

“I always come to this thing where... it's a problem with myself specifically, this idea of gimmick versus innovation... A new technology introduces a new kind of visual concept... but how do you go two more steps and, you're like, okay, this allows this, but what if we twisted it? What if we did something that we

didn't know if it was gonna work?”

Martin Madsen contrasted his workflow:

“Most of the time I know exactly what I need to create... I can't just indulge in creativity of how this can look. I have to create a workflow that creates this very specific thing... I found that AI comes into my workflow almost all the time.”

Technical and Budgetary Challenges

On the feasibility of high-resolution, long-form AI content, Martin was frank:

“Certainly not say it's easy... we have a big limitation with the context window of the size of the cards... I have to tile an image into smaller segments to be able to get that quality and resolution that we demand for high quality productions.”

Budgeting remains a challenge:

“It's insanely difficult... whatever tools we're building now are gonna be obsolete in half a year from now and we're gonna be working for two years. So how do you budget for that?”

Inspiration and Future Directions

On where they find inspiration, Paul said:

“It's the universities. I also just look at MIT and Berkeley, the white papers that they put out. That gives me the seed of the spark of an idea for something creative to use.”

Martin added:

“China's really interesting at the moment. It's a very fast moving game and it's very difficult to track.”

Drawing to a close, the fireside chat was a unique opportunity to bring together two leading figures from high-end storytelling and VFX for a candid audience dialogue. The format was highly active and well-received by the audience, setting a clear precedent for future editions.

The first "deep dive" session at Picture This 25 focused on Tallinn, Estonia-based Frost FX and its award-winning visual effects work for HBO's "The Penguin" series in 2025. The Estonian company earned both a Visual Effects Society Award and a Primetime Emmy for their contributions, demonstrating how specialized teams from very small countries and emerging film industry markets can achieve excellence in Hollywood production through strategic relationship development, technical innovation, and narrative-focused craft.

Frost Effects, represented at the session by CEO Marko Post and VFX Supervisor Eugene Bondar, completed 124 visual effects shots across four episodes during a six-month production period, deploying 21 artists. The technical scope encompassed complex CG rain systems, water simulations, Gotham City set extensions, and FEMA market environments.

Marko Post also challenged conventional wisdom suggesting companies in smaller markets should begin locally before expanding internationally. He questioned this limitation directly, emphasizing that mindset rather than geography represents the primary barrier to global competition.

He identified self-imposed limitations as the critical failure factor, arguing that accepting geographic constraints unnecessarily restricts potential. This perspective informed his strategy of targeting the United States market from inception, building capabilities and relationships to position themselves for premium projects regardless of their geographic origin.

Bondar also presented a philosophy distinguishing premium television work from effects-driven spectacle. He described the project as presenting complex, layered visual effects in every shot, but with effects designed to avoid drawing attention. The primary objective focused on creating atmosphere that supported storytelling rather than showcasing technical capability.

Manninen's session began with a fundamental technical overview of the principles of working of current language models and their critique. She argued that these systems are designed to generate the most statistically probable responses based on their training data, which inherently produces what she terms "generic shit" content—the most repeated, stereotypical, and conventional narrative patterns.

This limitation emerges from how AI models process information: they create multidimensional representations of word relationships from training data, then calculate token likelihood to construct "human-like" responses. The result is content that reflects the most common patterns in their datasets rather than original creative thinking required in film and creative arts.

The session went on to challenge prevalent assumptions about AI's knowledge base. Manninen argued that internet data, despite its volume, represents a severely limited subset of human experience which doesn't come close to the "out of the box experience" that drives great films. Most intimate, transformative, and emotionally complex moments—the raw material of compelling drama—remain undocumented. She noted that internet content is disproportionately produced by specific demographic groups and increasingly contaminated by AI-generated material, spam, and misinformation. This "shitty cup of data," as she described it, cannot substitute for the depth and authenticity of lived human experience that writers bring to their work.

Marta Krzeptowska (Producer, Head of International Development at ORKA Production Studio) opened the discussion by reflecting on the duality of her role, which bridges art house cinema and commercial production. She emphasized the profound impact of AI on the industry, noting, “We cannot control it. We cannot stop it… We have to talk and be aware." Her specific case study offered a candid exploration of the creative and technical challenges encountered when integrating AI into film production in the Polish context.

Marta Bródka (VFX producer and co-founder of Lumière de Cinema) provided an in-depth look at the technical advancements driving Polish VFX, especially for feature-length films. Bródka detailed the deployment of digital de-aging, aging, and face replacement technologies, which have become essential tools for maintaining character continuity and enhancing production efficiency. She described how her team has optimized workflows by leveraging internal adaptations of open-source AI models, enabling high-quality results while streamlining production timelines.

Joanna Andrecka (senior producer and post-production specialist at ORKA) recounted how AI has accelerated look development, enabling clients to select visual directions at the outset and significantly reducing the time required for concept and model creation. In the realm of 3D production, Andrecka’s team has utilized AI to generate fully textured models, saving substantial manual effort and delivering consistent, predictable results. She highlighted the use of LiDAR combined with AI for rapid, precise scene capture, reducing both on-set and post-production time. Andrecka also discussed the application of AI in rendering and set extension, which has proven invaluable for meeting tight commercial deadlines.

Importantly, she advocated for the strategic use of AI to automate repetitive post-production tasks, thereby freeing artists to focus on more creative and value-adding activities.

Throughout the session, a recurring theme was the necessity of adaptation and collaboration in the face of technological change. All Polish speakers stressed the importance of cross-border knowledge sharing, ethical engagement, and continuous investment in both technical and creative talent.

From the outset, Windelin set the goal of the session clear in his introductory words: while technology in film is always racing ahead, the true test lies in how these tools are applied in real productions, not just in theory. This year’s panel shifted the focus from technologists to top creatives, each presenting a Netflix-backed project that pushed the boundaries of what’s possible on screen.

Director Bobby Boermans and VFX supervisor Dennis Kleyn recounted the making of “iHostage,” a film inspired by a real-life hostage crisis in Amsterdam's Apple Store. The production faced a unique challenge: Apple refused access to its flagship store, the story’s central location. To account for that structural issue, the team decided to build a full-scale virtual production-driven replica of the store in an airplane hangar and used virtual production to blend real and digital worlds.

Production designer Niels Sejer shared how virtual production became integral to the creative process on “The Secrets We Keep.” Tasked with visualizing a world of privilege and existential tension, Sejer designed open, borderless sets that blurred the line between interior and exterior—a challenge for traditional sound stages, especially when the script demanded “eternal summer” in Denmark’s unpredictable spring.

The final case study was director Daniel Espinosa’s account of “The Helicopter Heist”, a complex Swedish project that essentially focused on using technology as a tool for creative freedom. Espinosa eschewed traditional storyboards and previs in favor of hands-on rehearsals and floor plans, using a gymnasium and iPhone capture to block out complex scenes. This approach, he argued, kept the process grounded in realism and performance, even as the production relied on virtual sets and advanced VFX.

To conclude, as Dennis Kleyn remarked during the panel, “You give back the environment to the crew… you create the environment and then the crew can use it as if it were a location.” This was echoed by all speakers who emphasized that the true value of virtual production lies in its ability to empower creative teams, foster collaboration, and maintain the emotional core of storytelling.