On June 18th, MiniMax, a trailblazer in AI technology, unveiled its latest video generation tool, Hailuo 02. This groundbreaking release is set to redefine the way creators worldwide approach video production, marking a significant milestone in the industry.

Hailuo 02 stands out for its ability to tackle intricate scenarios, such as gymnastics, which demand nuanced motion representation. The tool's launch aims to unleash the imagination of global creators, enhance video quality, and lower the barriers to entry for video creation. Behind the scenes, the MiniMax team has been tirelessly working on developing more efficient video generation model architectures, culminating in the introduction of Noise-aware Compute Redistribution (NCR).

The NCR architecture has revolutionized Hailuo 02, boosting both training and inference efficiency by 2.5 times. This enhancement allows the model to deliver enhanced performance with the same parameter count. To achieve this, MiniMax has expanded the model's total parameter count to three times that of its predecessor and increased the data volume by fourfold. As a result, creators can now generate high-quality video content at a lower cost.

Hailuo 02 is available in three versions: 768p-6 seconds, 768p-10 seconds, and 1080p-6 seconds. Thanks to its innovative architecture, Hailuo 02 offers a clear price advantage compared to similar products both domestically and internationally. Moreover, during early testing of the initial version, Hailuo 02 secured the second position in global rankings.

Looking ahead, MiniMax plans to continue rapid updates across various domains, including increasing generation speed, improving content preference alignment, and implementing advanced features such as image-to-video (I2V) capabilities. The launch of Hailuo 02 signifies a new chapter in video generation technology, garnering the attention and anticipation of creators worldwide.