The Rise of AI Video Models in China
In the fast-evolving world of Chinese short video platforms, two giants, Douyin (known globally as TikTok) and Kuaishou, are now quietly competing in the realm of AI video models. What began as a battle for user engagement and market share has transformed into a race to develop the most advanced AI video generation technologies.
Kuaishou's recent introduction of its AI video model, Kuaishou Kling, has set the stage for a new era of competition. This model, which was showcased at the World Artificial Intelligence Conference (WAIC) in 2024, allows users to generate realistic video content. The platform has already seen significant interest, with over 500,000 users applying to use it, generating more than 7 million videos.
Meanwhile, Douyin's parent company, ByteDance, is gearing up to reveal its own AI advancements. On July 19, ByteDance plans to share new developments in AI models, particularly in generating long and high-dynamic-range videos. The announcement follows the launch of ByteDance's World Models AI forum in Singapore, where the focus will be on cutting-edge techniques for continuous dynamic long video generation.
The Core Role of AI in Short Video Platforms
AI has become an integral part of the short video ecosystem, driving content recommendation, production, and understanding. As Kuaishou's Senior Vice President Gai Kun explains, the platform's business model heavily relies on AI to provide users with an immersive and engaging experience. Without AI, the short video business would not be sustainable.
Interestingly, Gai Kun previously worked at ByteDance, highlighting the interconnected nature of talent and technology between these two competitors. The short video platforms' reliance on AI underscores the importance of continuous innovation in this space.
Over the past decade, Chinese mobile internet has seen various attempts to capture user attention, from desktop browsers to weather apps. However, it is the short video platforms Douyin and Kuaishou that have emerged as dominant players, largely due to the integration of AI technologies. With over 600 million daily active users on Douyin and 394 million on Kuaishou, these platforms have become essential parts of everyday life for many in China.
The Future of AI Video Models
The release of OpenAI's Sora video generation model in February 2024 marked a significant milestone in AI development. Sora can create realistic, movie-like scenes based on simple prompts or static images, generating up to one minute of 1080p HD video. This breakthrough has inspired Chinese companies to develop similar capabilities.
Since May, several Chinese firms have unveiled AI video models comparable to Sora, including Shengshu and HiDream.ai. Kuaishou's Kling model, launched in June, has drawn considerable attention for its impressive video generation capabilities.
Kuaishou's AI efforts extend beyond video generation. The company is also exploring applications in video recommendation, content production, and understanding, leveraging large models like the ten-trillion parameter recommendation model SIM. These advancements are aimed at enhancing user experience and driving commercial opportunities.
ByteDance, on the other hand, has maintained a more secretive approach to its AI developments. Despite not being publicly traded, ByteDance has been actively pursuing AI model advancements. In August 2023, it launched the "Doubao" language model and the multimodal model BuboGPT. Douyin's Lark model has also been a significant development, gaining approval under the Interim Measures for the Management of Generative AI Services.
ByteDance has been building a comprehensive AI ecosystem, from foundational models to applications. The company has introduced AI-driven products like the chatbot Doubao, the AI Bot development platform Kouzi, and the AI learning companion Hippo. These initiatives reflect ByteDance's commitment to integrating AI across various aspects of its business.
As both Douyin and Kuaishou continue to invest in AI, the competition is expected to intensify. The focus will not only be on technological advancements but also on how these technologies can be commercialized. While Kuaishou has made significant strides in commercializing its AI capabilities, ByteDance's future plans remain more enigmatic.
The era of using AI to generate revenue is just beginning. According to a recent global survey by McKinsey, 65% of respondents reported frequent use of generative AI in their businesses, nearly double the figure from ten months ago. The service sector has seen the most significant increase in AI adoption.
Despite these advancements, the widespread application of AI models is still in its early stages. Many companies are focused on figuring out how to integrate AI into their operations effectively. The ultimate goal for Douyin and Kuaishou is to seamlessly incorporate AI into their platforms, enhancing user experience and driving innovation.
As the AI video model competition heats up, it will be fascinating to see how these two giants leverage their technological advancements to shape the future of short video content. The race is on, and the stakes are higher than ever.
