MiniMax’s Hailuo 02 Outshines Google Veo 3 in User Value

Post date:

Author:

Category:

MiniMax Unveils Hailuo 02: A New Era for Video AI Technology

In a groundbreaking move for artificial intelligence, MiniMax has unveiled Hailuo 02, an advanced version of its video AI model that promises significant upgrades in both performance and cost efficiency. This release showcases a cutting-edge architecture, Noise-aware Compute Redistribution (NCR), designed to enhance training and inference efficiency by a remarkable factor of 2.5.

A Closer Look at NCR Technology

Hailuo 02 employs the innovative NCR architecture, which optimizes how long video sequences are processed throughout the training stages. At the beginning of the training process, when artificial noise saturates the data, videos undergo maximum compression. Conversely, during the later stages—when the training videos clear significantly—the model operates at full resolution. Such strategic versatility allows Hailuo 02 to efficiently manage various complexities in video generation.



Improved Capabilities and Resources

Compared to its predecessor, Hailuo 02 boasts three times the parameters and utilizes four times more training data, a strategic enhancement that MiniMax claims leads to improvements in data quality and diversity. However, the company has not divulged specific parameter counts or dataset sizes.

Significant Gains in Complex Scenarios

MiniMax asserts that Hailuo 02 has marked improvements in its ability to handle complex prompts and simulate physical processes. According to the developers, this model is currently the only one capable of accurately generating intricate scenes, such as gymnastics routines.



Expansive Variants for Diverse Needs

Hailuo 02 offers three variants: 768p for six seconds, 768p for ten seconds, and 1080p for six seconds. This marks a notable upgrade; the previous model was restricted to 720p six-second videos, providing users with greater flexibility in content creation.

Hailuo’s Competitive Edge

In the Artificial Analysis Video Arena, a benchmark that assesses video outputs from competing AI models, Hailuo 02 claimed a commendable second place in the image-to-video segment. It narrowly trailed behind Bytedance’s Seedance and surpassed Google’s much-anticipated Veo 3 model. Notably, the current Veo 3 version does not enable audio support, a critical aspect for many video projects.

In user benchmarks, Hailuo 02 outperforms Google Veo 3, even though Veo also supports native audio generation. | Screenshot: THE DECODER

User Engagement and Growth

Since its initial demo launch in August last year, Hailuo has seen an astounding 3.7 billion videos created using the platform. MiniMax characterizes its rollout as “very random,” yet it quickly garnered attention and usage from creators around the globe.

Cost-Effective Solutions for Video Generation

Users can access the Hailuo 02 model via a web interface, mobile app, or API. For those utilizing the API, the cost to generate a six-second 768p video stands at $0.28, while a 1080p version is priced at $0.49. In comparison, creating an eight-second 1080p video with Google Veo 3 can run around $3, depending on the user plan.



Continuous Improvements Underway

MiniMax is committed to continual improvement, citing ventures to enhance video generation speed, stability, and expand features beyond the current text-to-video and image-to-video capabilities. Meanwhile, rival platforms like Runway already offer advanced features, including tracking shots, indicating that the competition is heating up.

The Launch During MiniMax Week

The launch of Hailuo 02 coincided with "MiniMax Week," a five-day event in which the company also unveiled an open-source language model, MiniMax-M1, complete with its specifications and technical documentation. Unlike Hailuo 02, specific technical details about its training architecture remain undisclosed.

Conclusion: The Future of AI Video Technology

With the Hailuo 02 model, MiniMax is not just upgrading its product but is setting the bar for future innovations in AI video technology. The significant enhancements in performance, cost-efficiency, and capability position it as a formidable contender in the video AI market. As the competition grows, users can expect even more dynamic features and improvements in the realm of artificial intelligence.

As the landscape of AI video generation continues to evolve, it will be interesting to see how MiniMax’s latest offerings shape the future of content creation.


By optimizing the discussion around MiniMax’s Hailuo 02, the article aims to rank well while providing readers with a comprehensive overview of the advancements in AI technology.

source

INSTAGRAM

Leah Sirama
Leah Siramahttps://ainewsera.com/
Leah Sirama, a lifelong enthusiast of Artificial Intelligence, has been exploring technology and the digital world since childhood. Known for his creative thinking, he's dedicated to improving AI experiences for everyone, earning respect in the field. His passion, curiosity, and creativity continue to drive progress in AI.