Sports Scenes Rendering Techniques

You will be amazed by how our sports games graphics have improved so much on portable devices as they seem to catch up with the console ones.

Gamers are able to enjoy uncompromised visual quality with full 4k resolution and up to 120 fps on mobile device, our game engine renders arena, 22 soccer players, and a crowd of 100k audiences with volumetric global illumination.

Analytical Volumetric Lighting, and high dynamic range lighting to make visual experience as genuine as possible, even more, benefit from multi-core parallel computing capabilities of latest mobile device, each audience on the stadium have their own behavior and action response to situations on the pitch.


Single Player AI & Player Group AI

Arena Intelligence is a data-driven AI system that runs virtual matches in Total Football. With Trajectory Engine, athlete's Kinematic Engine, and the Group Tactics Engine, our game takes players' on-pitch Intelligence to another level.

Virtual player’s behaviors are not only based on actual data of the players' physical strength, running speed, shooting accuracy, etc.. but also trained from tens of thousands of real games to learn players' unique habits and tricks, combined with user’s tactical settings and training path, our game brings endless freedom and authentic experience for users to explore, to compete, and to develop a great club.


Motion Capture and Character Motion System

Gala Sports has the largest motion capture venue in Asia for sports games based in Shenzhen.

We invited professional football players to record more than 2.000 hours of on-pitch movements and actions. After collecting all this data, we have put them together and built the MotionField action database, which is a data-driven action model that includes almost all possible human movements and actions that can happen on a football pitch.

This data-driven action system has become an important innovation in the sports gaming industry as it has greatly improved the quality and precision of the players' movements.

For example, when you dribble your opponent on Total Football, the video game's front-end search engine looks for the most suitable motion in the MotionField at a speed of 1 million times per second and displays it on your smart phone.


High Level Player Rendering

Since players' rendering is a crucial part of the video game, we simulate the movement of the player's muscles and jersey as realistically as possible by using a high-performance computer to reproduce the movement of bones, muscles, fascia, fat, skin, and jersey. Then, we use this data as an input for the neural network to reproduce these complex physical mappings on the screen.

The outcome is a high-quality muscle and jersey rendering, which could exceed PC and console high-end sports video games' display.


Automatic Character Modeling

Auto Portrait 3D is a players' 3D modeling pipeline developed by Gala Sports. This technology generates facial landmark and normal data by analyzing any player's picture, and back propagates and produces basic models with the Differentiable Renderer. Then with the using of Diffusion Modeling, we optimize and adjust each player's 3D model and mapping.

Finally, we use the Style GAN network to generate translucency maps and glosses maps with PBR+ subsurface scattering shader to render the 3D player in the video game.

The result is the world's largest 3D athletes database. For every registered FIFPro, NBA, and MLBPA player, we have created his photo-level 3D avatar.