Giant Network Group (002558.SZ) makes its first appearance at the Yunqi Conference, unveiling two self-developed large models with "game+AI" applications for the first time.

On September 19th, 2024, the 2024 Yunqi Conference opened in Hangzhou, with Giant Network Group (002558.SZ) participating for the first time with multiple new achievements in "game + AI." Two self-developed large models, GiantGPT and BaiLing-TTS, were launched, along with AI new technologies like Giant Momo Boundary and AIDIGIHUMAN, showcasing their leading-edge technology and application achievements in the field of artificial intelligence as pioneers in China in the "game + AI" sector. At the same time, the Giant Network Group AI Laboratory launched a new brand identity, with a rocket graphic composed of the letters G, A, and I, symbolizing Giant Network Group's exploration and prospects in the "game + AI" field. Since its establishment, the Giant Network Group AI Laboratory has been dedicated to achieving fully intelligent content production and creation, promoting innovation in gameplay, and the release of the brand identity marks a new stage in the development of AI business. The two self-developed large models in "game + AI" were launched for the first time, covering role-playing and speech generation. 2024 is considered the "year of large models landing" in the industry, with domestic large models experiencing rapid development and entering the stage of application in various industries. As one of the best testing grounds for AI technology, the gaming industry is accelerating its application. At the Yunqi Conference, Giant Network Group's self-developed role-playing large model, GiantGPT, and speech large model BaiLing-TTS made their first appearance, showcasing the latest application achievements of large models in gaming scenarios. Visitors to the exhibition could not only experience GiantGPT's role dialogue capabilities customized based on game IPs but also experience BaiLing's accurate generation of dialects and Beijing opera singing techniques with just one key input to feel the large model's abilities in both speaking and singing. GiantGPT is a large model focusing on the gaming industry, combining high-quality proprietary data and internet public data for training. It is optimized for basic capabilities such as role interpretation, scenario reasoning, long-term memory, etc., and its application effects have been verified in Giant Network Group's gaming business scenarios. For example, a personified question and answer system driven by large models has character traits and emotions, can fully understand player semantics, engage in natural and emotive conversations, providing players with a more realistic and lively experience. Another large model that made its first appearance is BaiLing-TTS, the industry's first speech large model supporting a mix of various Mandarin dialects. The Giant Network Group AI Laboratory has built a dataset covering 20 dialects, over 200,000 hours of Mandarin and dialect data, and employed multiple innovative technologies for large model training. BaiLing-TTS not only supports zero-sample cloni...