Smarter, safer, and better filmmaking in the age of AI

From left to right: Pietro Gagliano, founder & CCO, Transitional Forms; Hannah Elsakr, VP of Global GenAI New Business Ventures, Adobe; Ed Ulbrich, head of Strategic Growth & Partnerships, Moonvalley; Ryan Patterson, director, GRAiL; and Todd Terrazas, co-founder of FBRC.ai at “Spotlight: The Next Wave of AI Filmmaking: Smarter, Safer & Better” at TIFF.

We’re at one of those rare industry inflection points — a tectonic shift that feels a lot like the digital revolution of the 1990s and 2000s. Back then, tools like Photoshop prompted skepticism: People were concerned that the new applications would “destroy their craft.” Over time, industries embraced the new technology, new workflows emerged, and those same tools became indispensable in the world of creativity and beyond.

AI is on a similar arc. Its technical capabilities are advancing at a rapid pace, and it accelerates how fast creatives can iterate, test, and deliver new work. At the 50th Toronto International Film Festival (TIFF), I had the pleasure of participating in a panel on the next wave of AI filmmaking with three other leaders in AI and filmmaking — Ryan Patterson, director at Los Angeles-based AI studio GRAiL, Ed Ulbrich, head of strategic growth and partnerships at AI research lab Moonvalley, and Pietro Gagliano, founder and CCO of Transitional Forms. We discussed the ways AI is reshaping how stories are imagined, made, and shared.

Building on our conversation, this article explores the future of filmmaking, identifies essential elements filmmakers should preserve and insist upon, and explains how Adobe can support their journey — from the first scribble on a napkin to the final pixel on screen.

https://www.youtube.com/watch?v=R0ZGBHquy3Y

Ryan Patterson’s Dreamer, a short film about an aspiring filmmaker through the decades. From an endlessly creative kid in the 1980s to a struggling father in the 2010s, Dreamer explores the bumpy road to success and the power of inspiration. Made with Adobe Firefly, and produced in partnership with FBRC.ai, this film opened the panel “SPOTLIGHT The Next Wave of AI Filmmaking: Smarter, Safer & Better” at TIFF.

Shorten production cycles from idea to release

For many filmmakers, AI currently delivers the most value during early-stage ideation. As Patterson put it, “being able to get ideas out to people very quickly is a great opportunity, and it also enables creators to reach a larger audience because the filmmaking community is really rallying around this new process.”

That rapid iteration opens the sketchbook, allowing for more experiments, drafts, and chances to discover a surprising direction before a cent of the production budget is spent. It enables filmmakers to tell more stories that otherwise would never make it out of their notebooks.

Tools like Firefly Boards, for example, help creative professionals use AI to storyboard with far less friction, aligning on the vision and the overall look and feel before production begins. (For an example, see our case study of filmmaker Shona Dutta-Charlton.) Boards brings together our commercially safe Firefly models with partner models from Google, OpenAI, Runway, Luma AI — and now Moonvalley’s Marey.

From there, we’re very quickly moving towards an entire end-to-end workflow in which AI is supporting filmmakers in all phases of the production process: Generative AI tools like the Firefly Video Model accelerate pre-vis, complete scenes effectively, and unblock production problems (for example, a missing prop or a location that needs to be dropped because of budget restrictions).

Meanwhile, Adobe Premiere Pro and Adobe After Effects remain the industry standard for the final polish — now amplified by generative workflows that speed up iterations so that filmmakers can go from reference to rendered scene much faster.

Trust, rights, and responsibility

Powerful tools require responsible guardrails. Ulbrich, of Moonvalley, reminded us that many filmmakers are particularly concerned with transparency and trust.

That’s why our Firefly models are developed responsibly and ship with protections (for example, they are only trained on licensed content), while enterprise customers can work with custom, licensed models and defined guardrails.

This ensures that studios, talent, and distributors can fully embrace generative AI, use it at scale, and move from experimentation to confident adoption.

New forms, new economies, new audiences

We’re also watching the medium's boundaries expand. Gagliano of Transitional Forms, a studio-lab developing new forms of entertainment through creative machine intelligence, spoke to a bolder idea that I think will shape the next chapter: “The idea of dynamic, real-time media is coming.”

AI is going to change how we consume and how we participate, which is a really liberating idea for entertainment. In practice, this means putting the narrative or the brand in the hands of the consumer. Not just allowing fans to make edits or alternative endings, as we’ve seen before, but creating story worlds that aren’t fixed objects but hybrid living experiences — dynamic, interactive, and personalized in real time.

In this new paradigm, audience choices, agent-driven characters, and multi-agent systems will alter narrative arcs and create entirely new forms of storytelling, distribution, and monetization.

There’s also a shift in who can make films. With fewer gatekeepers and more accessible tools anyone with a phone and a vision can create something amazing all on their own. That democratization is exciting, but it also raises questions about IP, compensation, and creative authorship that the industry must address.

The right commercial and technical frameworks will let creators monetize, studios train custom models on their archival catalogs (turning decades of films, assets and style references into controlled AI tools that speed up new productions and unlock new ways to modernize franchises) , and audiences participate without undermining the livelihoods of those who built the originals.

Embrace the unpredictability

AI tools are becoming more precise, but sometimes unexpected outputs — the so-called “hallucinations” generated by AI that people often worry about — can actually be creative fuel if handled thoughtfully.

As Gagliano explained, the best experiences mix predictability with surprise, and his team at Transitional Forms even created a weirdness policy to use unexpected outputs as a jumping-off point for their creativity. Ulbrich added that powerful models sometimes reveal results that haven’t been planned but turn out to be creatively useful. These accidental discoveries can become new tools in a filmmaker’s toolbox.

In a different medium, furniture designer Norman Teague, who used Firefly to reimagine MoMA’s design collection, embraced hallucinations because they put two things together that he wouldn’t have thought of otherwise — for example, in his record turntable armchair that was on display. He called Firefly GenAI a search engine for his imagination. He was able to imagine, generate, and realize many more of his ideas than would previously have been possible.

A pragmatic, creator-first future

As Ulbrich, Gagliano, Patterson, and several others noted at TIFF, the tools are exciting; the work now is to make them trustworthy and enhance them further for a commercial context. And so, the next wave of AI filmmaking involves reshaping the toolkit and the relationship between maker and audience.

It will be “smarter” when it augments the craft; “safer” when trust and rights protection are built in; and “better” when it puts creators — not algorithms — at the center.

The practical advice for filmmakers is this: Experiment early, preserve your IP, and demand transparency. If you’re a director, producer, or artist, be clear on what you want to own and how you want your work to be used — and include that in contracts now. If you’re a studio, consider how custom models can unlock new revenue from existing vaults without eroding the underlying IP value.

One last operating idea I keep returning to: “Fix it in pre.” When AI helps you solve problems earlier in the process, teams can deliver better, faster, and with fewer compromises.