Shanghai-based AI firm MiniMax has officially launched its H3 video generation model, marking a significant entry into the competitive field of multimodal AI. The new model can process text, images, video, and audio to generate videos up to 15 seconds long with 2K resolution and native stereo sound. This launch positions MiniMax as a direct challenger to prominent Chinese rivals like ByteDance with its Seedance 2.0 and Kuaishou with Kling 3.0.

H3's capabilities extend beyond basic generation, allowing users to edit existing content and transfer movements from one video to another using various input formats. MiniMax plans to release H3's model weights within days, embracing an open-weight approach increasingly adopted by tech giants. This strategy aims to foster broader adoption and customization of its underlying system, potentially accelerating innovation within the developer community.

The model is specifically designed for commercial applications, including advertising, e-commerce, product design, and games, with MiniMax claiming that generating 2K video will cost less than one-third of mainstream rival products. Furthermore, in a strategic move reflecting China's broader technological independence goals, H3 is engineered to work with several Chinese-made chips. This focus on domestic hardware integration highlights a growing trend within the Chinese AI sector to reduce reliance on foreign semiconductor technology.

The release of H3 intensifies the race among Chinese video-model developers, demonstrating rapid advancements in the region's generative AI capabilities. The ability to produce sophisticated clips from diverse inputs is becoming a critical differentiator, pushing the boundaries of what's possible in AI-powered content creation. This competition is not just about features but also about cost-efficiency and strategic hardware partnerships.