FIFA 22 uses machine learning to make player animation more realistic

FIFA 22 will mark a turning point in animation technology. The 29th installment of Electronic Arts’ famous football simulation will be released on October 1. One of the novelties of this vintage concerns its design. As usual, EA filmed motion-capture players to develop the game’s animations. But as the game’s production manager at the Polygon specialist site explains, the most interesting thing is what didn’t. not been registered.

Better tactical realism
The company has been working for three years on a machine learning algorithm, which learns data from motion capture sessions to be able to recreate animations in real time. « How the players approach the ball, how many steps they take to get to it, the length of their strides, at what angle, what acceleration, etc. », he explains. « It is quite simply the beginning of machine learning which takes over from animation ». As a result, greater tactical realism thanks to better management of offensive and defensive movements. For FIFA 22, EA captured the footage of 22 professional players in a 90-minute match, which produced 4,000 new animations.

Thanks to this technology, called HyerMotion, the game is able to use the 8.7 million images from motion captures to recreate others in real time, depending on the game situation. realism in animation transitions. According to the publisher, the change will be particularly visible in the collective movements of players on the field, especially when the blocks go up or down.

The best-selling game in 2020
This technology could be put to good use in other EA games, but the work of interpreting movements would have to start from scratch for another sport. In 2020, FIFA 21 was the best-selling video game in France. It has passed 1.3 million copies according to the French Syndicate of Video Game Publishers.