Pixar
While the debate over the use of AI is gaining traction across the entertainment industry, the technology has quietly aided animation and visual effects staff for years. Some of the most amazing visual images are possible when you ask craftsmen to do things that were previously thought impossible.
When director Peter Thorne wanted a character based on the elements of fire, water, air, and earth in his new film Elemental, VFX supervisor Sanjay Bakshi and the Pixar team decided to make the process smoother. focused on AI. The character’s appearance had to be adjusted to match Sohn’s vision.
“We used AI for a very specific kind of problem and used a machine learning algorithm called Neural Style Transfer,” says Bakshi. “Our animation is very scrutinized. So many reviews for every shot. He’s been through a cycle and really handcrafted by our animators so we can apply machine learning in its current form. There are not many places.
“But there’s one problem with Elemental: running a fire simulation on top of the character to make it feel like it’s burning. Then the flames themselves go through a very realistic pyro-sim. It’s a fluid simulation and a real temperature simulation, so the flames it produces are very realistic.We needed a way to stylize the flames themselves.As you can imagine, the styling of the simulation is It’s not an easy problem.It’s very temporary.It’s always changing.And that’s the beauty of fire.It’s always different and it’s very interesting to watch.The art of stylizing fire is so Not many, but we found a technique called Neural Style Transfer, and we used it, and it was really the only viable solution.”
Gavin Kelly, founding partner of Piranha Bar, a Dublin-based animation and VFX house, also said AI will be used more and more as animators and content creators seek to push the boundaries of visuals. We believe that it is a technology that will become
“On the far end, we’re not quite there yet, but you just shoot something and tell the AI what you want it to be in terms of performance capture,” says Kelly. “So, capturing a performance is very complicated. You put together the animation, you put the facial headset in place, you talk to the software that talks to your hands, your body and everything else. They’re all different parts, Let’s talk it all together.To create this pipeline is very complicated and there will be a lot of troubleshooting along the way.So at this point it’s just that AI motion capture solutions have existed in the past No doubt about it, we saw them before they became terrible and weren’t production ready and now you can rotate the camera and the AI will take care of that and it’s robust So we are very close to production readiness, and it will be solid and very compelling.”
For Bakshi and his team, AI still requires careful tuning by artists and VFX staff to achieve the desired visuals. Nothing can be taken for granted.
“The guy who worked with us on AI was Jonathan Hoffman, and he described it as like throwing fish into a tornado and trying to get sushi from a machine learning algorithm,” Bakshi said. laughs. “So you can type in what you want and you might get something really beautiful, but you still might not get what you want from the animation that comes back.”
