China's MiniMax Launches H3 Video Model, Intensifying AI Race

Chinese artificial intelligence firm MiniMax on Friday launched a new video-generation model, a move that heightens competition in the rapidly advancing generative AI sector dominated by domestic tech giants.

The Shanghai-based company, considered one of China's leading "AI tigers," announced the release of its H3 model, which is designed for a range of commercial applications.

H3 Model Capabilities

According to the company's announcement, the H3 model possesses several advanced features aimed at both content creation and editing. MiniMax stated the model is capable of generating high-quality video clips from various inputs.

Key features include:

• Video Generation: Creates clips up to 15 seconds in length at 2K resolution.
• Multi-modal Input: Processes text, images, video, and audio to generate content.
• Native Audio: Includes native stereo sound with the generated video.
• Cost Efficiency: The company claims the cost to generate 2K video is less than one-third of that of mainstream competitors.

Ad

In a significant strategic decision, MiniMax plans to release H3's model weights within days. This open-weight approach will allow users to download and customize the underlying system, a departure from the proprietary models common in the video generation market.

Competitive Landscape and Strategy

The launch of H3 places MiniMax in direct competition with major players like ByteDance, which released its Seedance 2.0 model earlier this year, and Kuaishou, which developed Kling 3.0. Founded in 2022, MiniMax is a well-funded startup that went public in Hong Kong in January, as reported by Reuters.

The company is targeting commercial use cases such as advertising, e-commerce, product design, and gaming. Furthering China's push for technological self-reliance, MiniMax said H3 was designed to be compatible with several Chinese-made semiconductor chips.