Autodesk's Flow Studio isn't another flashy one-click generator designed to spit out viral clips for socials, sorry Will Smith stans. Instead, Flow Studio is quietly doing something harder: translating the messy, tactile craft of filmmaking into a set of AI tools that think in cameras, characters, and lights, not just pixels.
That distinction came through repeatedly during my conversation with Nikola Todorovic at Autodesk University 2025. The founder of Flow Studio is now helping steer his AI software inside Autodesk, where the start-up's ambition meets the scale of a major tools company, and a new tiered pricing model that includes a 'freemium' license.
"I read it this morning," Nikola deadpanned when reminded of the news cycle, a modest opening that belied the scale of what the team has been building. The point he kept returning to was simple: generative video models are great at creating images, less good at giving filmmakers the kinds of levers they actually need.
As he put it on stage and again in our chat, the trouble with the current crop of text-to-video tools is that "it's very hard to manipulate them [...] I can't prompt the camera to move to the intro I want, right? I can't prompt the performance, or even, animation - the body, pose - into these spaces."
Flow Studio addresses that by approaching the problem as a multi-modal one: separate models for character, camera, environment, and lighting, so filmmakers can work with components, the same mental model used in traditional CG. "Everything we build, all the models we have, are meant to separate the world, just like we do in traditional CG," Nikola explains. That separation is Flow's secret weapon: don't try to replace the filmmaker's vocabulary, augment it.
That philosophy isn't theoretical. Flow started life as an integration into Maya: "We built this integration, this tool, kind of add-on to help prepare your characters in Maya," Nikola says, and transferring rich data back into pipeline tools has been a focus since the company's earliest partnership with Autodesk. The idea, he said, is to deliver not just a rendered video but the metadata that makes that render usable inside a 3D scene. "Our tool is kind of an in-between step to provide you all this data so you can go and really add to it, you know, the camera, character, and lighting, etc."
That emphasis on data, on preserving a 3D scene from video, is what makes Flow Studio feel less like a novelty and more like a production tool. Nikola emphasised that the goal isn't to obliterate traditional production: "I'm always a big believer of needing a traditional way of editing things, because you can't do everything with a prompt. And then, I don't want to get rid of shooting on set."
This brings us to one of the most interesting cultural debates surrounding AI tools in film: who benefits? Nikola's origin story is telling. He grew up in the former Yugoslavia, moved to the U.S. at 20, worked as a VFX artist and supervisor, and learned how brutally hard it can be to break into filmmaking. "I was always really frustrated by how hard it is to break in and how difficult it is to get film funded," he says. Flow Studio was partly born to solve that problem: "We were writing a film about robotics, and then we realised we didn't have the budget for it. So [we decided to] start looking into this tool."
The software's adoption reflects that mission: despite its ambitions, Flow is already being used in larger studio contexts, but Nikola's excitement lies with indie creators. "Personally, I'm more excited about indie and smaller production companies. But obviously big studios have been using [it]," he says. The promise is both pragmatic and cultural. It lets filmmakers iterate faster, produce more shorts, and learn by doing, rather than waiting years between productions.
Nikola is also candid about the limits of automation. Even Flow's clearest early wins - "easy VFX shots" - still need cleanup and manual editorial work. "The idea is, the harder the VFX shot is, the more clean up you gonna have, and the more manual editing you're going to do," he says. "We never wanted to build something that would automatically do everything. I don't think that's possible."
That humility sits beside a sharp optimism. Nikola expects a gradual, not a meteoric, shift, one similar to other technological shifts in film history. "I do think it will happen gradually... a lot of these tools for generative AI are targeting social media first," he notes, and predicts the pattern we've seen before: advertising and music videos, then feature work adapts. He also believes new voices will emerge: people using the tools to discover storytelling forms we haven't yet imagined.
Inside Autodesk, that connection to the broader ecosystem matters. Nikola stresses that Flow's roadmap is merging with other content tools, particularly MotionMaker and FaceAnimator. "Our roadmaps are merging. So it's not that we're completely separate, but as we develop something where we see, 'oh, we're developing this, oh, you're developing this, great, let's merge efforts,'" he teases, specifically name-checking MotionMaker as a standout update.
There are inevitable anxieties, the fear that synthetic production will strip away the magic of shooting on set is "very much justified," Nikola acknowledges, but his position is pragmatic: tools should expand the field of who gets to tell stories, not flatten the craft. "I really hope we don't turn into just synthetic movies where actors are replaced by their likenesses. I think, really think, that would be a failure."
Autodesk's Flow Studio is, then, less a threat and more a proposition: what if generative tools learned to speak the language filmmakers already use, delivered the data pipelines needed, and let creators practise the craft more often? If Nikola's conviction holds, the next decade won't be about whether AI replaces filmmakers; it will be about how the tools raise the bar for what those filmmakers can imagine and build.
