Tencent Details Generative AI Use in Game Production at CEDEC2026

Tencent Japan's Yuejiang Li took the CEDEC2026 stage to explain how generative AI can be woven into the entire game production pipeline, from design to QA. The talk, titled "How Far Can AI Change Game Production?", drew on Tencent's own practices and cited a GDC report: 36% of developers already use generative AI, while 52% of industry professionals view its impact negatively.

Li argued the industry does not want AI that replaces creativity, but rather AI that handles repetitive and tedious tasks, letting teams focus on essential creation. Tencent treats generative AI not as a standalone tool but as something that creates value only when integrated into each process. That integration requires a common infrastructure supporting connections to multiple AI models and usage monitoring.

In design, Tencent introduced the creative support tool Miora. On a freely expandable canvas, designers describe their vision in natural language, and an AI agent breaks it into tasks, selects models, and generates images. Those images can become 3D models with rigging and motion, then checked as videos with camera work. People compare and select results at each stage, giving correction instructions on the spot. Miora also accumulates designers' thinking and project rules as AI memory, carrying previous policies into new content to keep consistency across a project.

The lecture also covered production bottlenecks, such as global illumination lighting. Conventional baking has high computational load, making lightmap generation a bottleneck. Tencent's approach uses AI to streamline that process, though details on the specific tool were cut off in the report.