QianYing: Giant Network's Audio Game Generation Model 137

"QianYing" is an audio game large model launched by Giant Network, which includes the game video generation large model YingGame and the video dubbing large model YingSound. It achieves the generation of audio interactive game videos, marking a technological breakthrough in the field of ai games. YingGame is jointly developed by Giant Network AI Lab and Tsinghua University's SATLab, supporting interactive control of diverse character actions, allowing users to customize game characters.

Main features of QianYing

• YingGame: As a game video generation large model, it can create open-world game videos, support controlling character actions through text, images, or traditional keyboard and mouse operations, and allow users to customize game characters. YingGame's strengths lie in diverse character movements and personalized design, as well as physical simulation characteristics that do not require traditional 3D modeling and rendering.

• YingSound: A video dubbing large model specifically designed to generate matching sound effects based on video footage, including complex ambient sounds and character dialogue, without the need for manual prompts, directly extracting semantics from video content for high-precision audio generation. YingSound's generalization capability makes it applicable to various video types, from real-life scenes to anime and games, enhancing game immersion.

QianYing Model Technology:

• YingGame: Jointly developed by Giant Network AI Lab and Tsinghua University's SATLab, supporting interactive control of diverse character actions.

• YingSound: Jointly developed by Giant Network AI Lab, Northwestern Polytechnical University's ASLP Lab, and Zhejiang University, with exceptional time alignment and video semantic understanding capabilities.

• Open-world games: YingGame can generate open-world game videos, achieving interactive control of character actions through text, images, or keyboard and mouse operations.

• Sound effect generation: YingSound can generate complex sound effects matching the visuals, achieving audio game videos.

• Technological breakthrough: The QianYing model demonstrates the feasibility of generating game content through text descriptions.

QianYing Co-Creation Plan

• Plan launch: Giant Network launched the "QianYing Co-Creation Plan", opening up cooperation in model technology, applications, and original game projects.

• Cooperation areas: Including cooperation and co-creation in model technology, model applications, original game projects, and other aspects.

• Application implementation: Accelerating the implementation of "Game+AI" technology, promoting the transformation from cost reduction and efficiency improvement to disruptive innovation.

• Talent recruitment: The plan also includes talent recruitment to support the continuous development of technology.

• Openness: The plan aims to accelerate the maturity and application of technology through open cooperation.

Giant Network launched the "QianYing" audio game generation large model at the 2024 China Game Industry Annual Conference, demonstrating Giant Network's significant technological development in the "AI Games" field.