发布时间: 12/30/2025

From December 15 to 18, the 18th ACM SIGGRAPH Asia conference and exhibition lit up the Hong Kong Convention and Exhibition Centre (HKCEC). This premier event in computer graphics and interactive tech zeroed in on the theme 'Generative Revival,' exploring how AI is transforming creative industries. Picture over 7,000 enthusiasts and 600 speakers from around the globe, with heavyweights like Tencent, Huawei, and Adobe unveiling their freshest AI-driven graphics breakthroughs.

Tencent neural graphics and generative AI workshop scene at SIGGRAPH Asia 2025

Tencent stole the show with workshops on 'Neural Graphics and Generative AI,' diving into cutting-edge 3D perception world models and tracing the journey from research papers to massive-scale products. Tencent Games dropped 9 accepted papers and 3 spotlight sessions, zeroing in on how AI reshapes asset creation and engine rendering, sparking lively chats with top AI and graphics pros worldwide.

Tencent Games AI asset creation and rendering discussions at SIGGRAPH Asia 2025

AI Revolutionizing Game Production Pipelines for Superior Content Output

Tencent Games' tech squad presented the paper 'Imaginarium: Vision-guided High-Quality Scene Layout Generation' in the Technical Papers track. It tackles the headache of time-sucking layouts for non-core zones in open-world games, like bustling streets or wild foliage, via a smart visual-guided 3D scene generation approach. Tencent expert Zhu Xiaoming demoed Imaginarium as an AI system that groks text prompts, thinks like a pro designer, and spits out style-matching, logically placed 3D scenes—perfect for fleshing out vast maps so artists focus on killer creative bits.

Imaginarium AI 3D scene generation system demo by Zhu Xiaoming at SIGGRAPH Asia 2025

Zhu Xiaoming's talk: 'Imaginarium: Vision-guided High-Quality Scene Layout Generation'

What sets it apart? It doesn't just mimic placement; it reasons 'why' behind layouts, grasping scene logic and story vibes for that human-like 'slow thinking' magic.

AI-generated cozy living room from text prompt using Imaginarium at SIGGRAPH Asia 2025

Here's a sample output from the prompt: 'a cozy living room with comfy armchairs, gallery wall, and stylish coffee table.'

Building on their Gamescom reveal, Tencent's VISVISE team rolled out the industry's first full AI 3D animation pipeline. Lead Zeng Zijiao broke it down in 'Breathing Life into Geometry: Full AI Pipeline for 3D Character Animation,' covering upgrades to skeleton generation, smart skinning, 3D animation, and smart inbetweening. These plug straight into game pipelines, ramping up efficiency, with live demos at their booth showing seamless 3D character animation workflows.

Zeng Zijiao speech on VISVISE full AI 3D animation pipeline at SIGGRAPH Asia 2025

Zeng Zijiao's presentation: 'Breathing Life into Geometry: Full AI Pipeline for 3D Character Animation'

It's already powering nearly 100 titles like Peacekeeper Elite, Honor of Kings, PUBG Mobile, Teamfight Tactics, and League of Legends: Wild Rift, boosting production speeds by over 8x.

VISVISE AI 3D character animation workflow diagram at SIGGRAPH Asia 2025

VISVISE workflow demo

VISVISE smart skeleton rigging generation process in action at SIGGRAPH Asia 2025

Smart skeleton rigging in action—generating now

Photon Studio Group's Wu Kui and Zheng Zhongtian shared 'Automatic Hair Strand Extraction via Differentiable Rendering' in Technical Papers, covering asset breakdown, preprocessing, and rendering tricks to cut redundant art tasks without skimping on quality.

Wu Kui and Zheng Zhongtian on differentiable rendering hair extraction at SIGGRAPH Asia 2025

Wu Kui and Zheng Zhongtian's talk: 'Automatic Hair Strand Extraction via Differentiable Rendering'

Technical demo of hair strand extraction tech visuals at SIGGRAPH Asia 2025

SPGen spherical projection 3D shape generation examples at SIGGRAPH Asia 2025

Old-school multi-view 3D modeling often stumbles on geometry mismatches and tricky topologies. Tencent's Weikai Chen and team introduced SPGen, a spherical projection method for single-image 3D shape generation ('Spherical Projection as Consistent and Flexible Representation for Single Image 3D Shape Generation'). It projects surfaces onto a bounding sphere for flawless consistency, nailing complex geometries with diffusion priors, slashing GPU needs (just 96GB on two cards), streamlining game dev for smarter, complex 3D modeling.

Meanwhile, Du Xingyi's crew tackled generalized sweep volumes with 'Lifted Surfacing of Generalized Sweep Volumes,' delivering precise math for self-intersection-free, watertight surfaces. This amps up accuracy for wild geometries and topologies, speeding up virtual worlds and props in games.

AI + Game Engines: Supercharging Rendering and Immersive AV Experiences

Tencent Games flexed AI across lighting, animation, physics, geometry, and real-time rendering, paving the way for deeper immersion and smoother gameplay.

Graphics evolution keeps amping game visuals, with in-game AI rendering stealing spotlight. Tencent's rendering lead Li Chao unpacked MagicDawn—their cross-engine lighting solution—in 'Frontier Rendering Tech and Game Practice: Breaking AV Limits for Next-Gen Experiences.'

Li Chao unveiling MagicDawn cross-engine AI lighting solution at SIGGRAPH Asia 2025

Li Chao on-site: 'Frontier Rendering Tech and Game Practice—Breaking AV Limits, Next-Gen Experiences'

MagicDawn handles dynamic day-night cycles cross-engine, uses AI to shrink data and boost speed, slashing bake times from days to hours—even delivering AAA shadows on mobiles. It nails global illumination pains with end-to-end scalability, powering Escape from Tarkov: Dark Zone, infinite modes, Wuthering Waves, and Lok Kingdom Worlds.

MagicDawn dynamic lighting effects before-after comparison at SIGGRAPH Asia 2025

Intelligent real-time rendering showcase with AI optimizations at SIGGRAPH Asia 2025

Smart rendering showcase

On animation, Tencent Magic Square Group's Xilei Wei, Lang Xu, Yeshuang Lin, and Zhejiang Uni teammates presented 'Ultrafast and Controllable Online Motion Retargeting for Game Scenarios.' Their semantic-aware geometric rep solves retargeting across body types in 0.13ms per frame, nailing contacts and weapon plays for vibrant massive worlds.

Ultrafast motion retargeting algorithm demo for game characters at SIGGRAPH Asia 2025

For geometry, Photon experts like Wu Kui collabed with Zhejiang Uni on 'RL-ACD: Reinforcement Learning-Based Approximate Convex Decomposition,' compacting complex scenes/characters for scalable interactive realms. Wu Kui also co-led 'Kinetic Free-Surface Flows and Foams with Sharp Interfaces,' mastering efficient, stable fluids with foam details.

Facing high-fidelity/high-fps hurdles, Yuan Yazhen's team offered predictive sparse coloring for frame extrapolation—rendering only busy spots under tight budgets for stable, detailed high-fps. For VR/AR, 'StereoFG: Generating Stereo Frames from Centered Feature Stream' alternates low-res eye renders, AI-filling the other for 4x stereo efficiency.

ACM SIGGRAPH chair Chen Baoquan from Peking U nailed it: this conference spotlights '3D graphics' pivotal role in AI and emerging industries.' Games, blending 3D, physics, and real-time interaction, are AI's playground.

Tencent Games has woven AI into games since 2016, tackling R&D woes while debuting smart assistants and AI hounds in Honor of Kings and Peacekeeper Elite. Leveraging games' realism and interactivity, frontier AI lands fast, turning billions of players into real-world testers. These SIGGRAPH shares? They'll morph into the next wave of tangible, interactive game magic players crave.