How AI May Solve Football's Most Controversial Rule At This FIFA World Cup

In a quiet demo room in Zurich, far from the noise of stadium crowds, football’s future was being rebuilt, particle by particle. I was there because Lenovo and FIFA have teamed up ahead of the FIFA World Cup 2026. And while most partnerships talk about “fan experience” in vague terms, this one gets surprisingly specific -- fixing one of football’s most frustrating elements, aka the offside decision.At the centre of it all is Valerio Rizzo, AI Senior Manager & Solution Architect (3D Digital Avatars) at Lenovo who is a neuroscientist turned AI architect, leading the development of what can only be described as the next evolution of VAR (Video Assistant Referee) visuals. Not the decision-making itself, that remains intact but how those decisions

are shown to the world. And that distinction matters more than you’d think.Problem With Grey PuppetsIf you’ve ever watched a VAR offside replay, you know the visuals. Simplified player models, awkward movements, and a lack of realism. They do the job, technically but they don’t feel right. Rizzo mentioned during the session, “the current system creates friction.” It’s not that the decision is wrong, it’s that the way it’s presented makes people doubt it.

When you see a low-detail, almost video game-like animation deciding the fate of a goal, your brain instinctively questions it. It doesn’t look real, so it doesn’t feel trustworthy.The goal isn’t just better graphics, it’s the accuracy. Instead of generic avatars, the new system creates what Rizzo calls a “Digital Twin“ of each player, which is an actual 3D reconstruction of the player’s body, proportions, face, and even hair. This means when an offside is shown, you’re not looking at a placeholder figure. You’re looking at that exact player, in that exact moment, reconstructed in 3D space. And suddenly, the decision feels a lot more real.How It Actually WorksCreating one avatar takes around three hours. That doesn’t sound fast, until you consider the scale. Over 1,200 players need to be scanned and processed. To make that possible, Lenovo is deploying 28 scanning setups across team base camps during the tournament.Traditionally, creating detailed 3D models requires hundreds of cameras and massive setups, something you’d expect in Hollywood studios. Lenovo’s approach to the entire thing is very different. They managed to put 36 cameras inside a chamber, where players can enter and just get a scan.Each scan itself takes just milliseconds, literally, I got one too. The heavy lifting happens afterward in the cloud, where AI models reconstruct and refine the avatars. The process can run across multiple GPUs and systems simultaneously, cutting down overall time.

Instead of relying heavily on hardware, they lean into AI. The system uses a technique called 3D Gaussian Splatting. In simple terms, it takes multiple images of a player and converts them into a cloud of tiny particles. Each particle carries information such as colour, position, depth and when stitched together, they form a highly accurate 3D representation.One of the biggest challenges in 3D reconstruction is something you’d never think about -- hair. It’s not a solid object and every single player has about lakhs of tiny strands. That’s why many 3D characters look slightly “off.” But with this particle-based approach, even fine details like hair strands can be captured more naturally. It’s not perfect yet, but it’s a big step forward.Then there’s the kit problem. If you scan a player wearing a jersey, that jersey becomes part of the model. So how do you change it later for different matches, numbers, or kits? The answer is segmentation. An AI model separates the body into parts -- skin, clothing and boots, so each element can be modified independently. That means you can swap jerseys, change numbers, or adjust colours in real time without rebuilding the entire model. Think of it like Photoshop, but for 3D humans.An Upgrade, Not ReplacementOne of the smartest decisions here is what Lenovo didn’t do. They didn’t try to rebuild VAR. The existing Semi-Automated Offside Technology (SAOT) already works well. It tracks player positions using cameras and reconstructs skeletal movement with high precision.This new system simply adds a layer on top of it. The skeleton data remains the same. What changes is what you see attached to that skeleton. Instead of a generic figure, the system maps the real player’s digital twin onto it. The movement stays accurate, but now the visuals match reality. So, it’s not a replacement to anything that you see, it’s an upgrade.The Bigger PictureBesides the scenic views, what stood out most in Zurich wasn’t just the tech, it was the intent behind it. Tech was made to solve our problems, and this one does that too. While this project is built for FIFA, the implications can go much further. The same technology could be used in Gaming and VR, films, fashion and even medical imaging. Once you can create accurate digital humans quickly, the applications are endless.Football just happens to be the proving ground and, in a sport, where a single offside call can change everything, that’s no small thing. In 2026, when fans watch an offside replay, they might not notice the underlying AI, the particle clouds, or the scanning infrastructure. What they will notice is this - it almost looks real and sometimes that’s all it takes.