Michel Ancel on what’s next for Ubisoft Montpellier’s AI-driven development tech


Ubisoft Montpellier is working on an answer to the colossal budgets, teams and studio spaces that characterise modern game development. The publisher’s smallest self-contained team gambled on Beyond Good & Evil, turned a quick movie cash-in into a passion project with King Kong, and then sold enough copies of 2D platformer Rayman Origins to make a sequel, Legends. Montpellier’s budgets are slim, its team is tiny and the studio is a house on a hill, but with Rayman the studio has built a foundation upon which bigger, broader games will be made.

“Compared to Assassin’s Creed [the team is] pretty small,” studio head Michel Ancel says. “Of course, when you work with a small team, the big advantage is that the potential of everyone can be seen in a very direct way, as long as your way of making the game is very flexible. So that’s really the key – if you don’t have this flexibility, you could lose the potential of everyone in the team.”

UbiArt Framework was Montpellier’s first step in producing an art-intensive game with flexibility and speed. All an artist needs is a silhouette, and an animator can have it moving in-engine within minutes. “I think UbiArt, when you see it the first time, is really amazing,” Rayman Legends’ lead game designer Emile Morel says. “You don’t even have to pause the level. You can put an object or an enemy where you want, even while you’re playing. It gives you the opportunity to try new things and do things very fast and improve all the time. All these elements make the iteration process very fast and easy to use.”

Originally developed for 2010’s Rayman Origins and part-funded by the French government for its contribution to the arts, UbiArt Framework was the result of years of research and development. A trip to New Zealand’s Weta Digital, responsible for the visual effects on Avatar and The Hobbit, helped. “We visited the studios in New Zealand,” Ancel recalls. “We discovered they were working closely with the programmers in order to avoid too much repetitive work done by humans. I think it’s very interesting to see how much you can integrate and put artificial intelligence in the tools so you can really order the computer to do things. It’s a very interesting way of working, to see where and how humans are involved in creation [and where the computer takes over].”

Could these methods of production for a 2D game carry over into 3D? More tantalisingly, could they carry over into Ancel’s long-awaited labour of love, Beyond Good & Evil 2? “Of course,” Ancel says. “There’s not so much difference between the two because, as soon as you start the process of sharing human [development] with the computer, you can translate that to more complex games. We have done research and it’s something we are working on, but I can’t mention too much about that.”

Offloading work to UbiArt in order to save on manpower is efficient and cost-effective. Better still, it frees the studio’s artists and designers to work on the parts demanding a human touch – art, level design, audio – while the computer’s logic is committed to UV mapping, dynamic lighting, collision detection and so on. The engine intelligently couples an ‘invisible wall’ to the level’s parameters to handle collision detection without ever being touched by a designer. A small change to level geometry in other games can create a cascade of problems; in UbiArt’s system, geometry and collision detection are one and the same.

Those techniques, Ancel says, will carry over to large-scale open-world games such as Beyond Good & Evil 2. “For sure, we’re doing research that could help even for bigger games,” he says. “We believe we can achieve big games even with this size [of studio]; more could be done by the computer and we can focus on the really important things done by real people.

“The content and the quality of the content will be different. That’s something we’ve learned with our tools on Rayman Legends; in fact, those tools improve the quality. It was also funny to see how many people in the team can create content with the tools. Sometimes in [this business] you say, ‘Oh, this veteran is better than this young guy,’ but what if the veteran is only better at the technical things? When those two people are using tools that are easy to use, technical things don’t matter; it’s more the imagination and ideas that are very important. But we need to practise and use the tools, and now that Rayman Legends has been finished, we are going to move back onto triple-A games with those tools. I think it could be very interesting to talk about it in a year, and to see how far we’ve come.”