(Yicai) April 10 -- HappyHorse, an artificial intelligence video generator that has gone viral after storming global rankings, was developed by Alibaba Token Hub, a business group set up last month by Chinese internet giant Alibaba Group Holding to bring a laser-like focus to its AI strategy.

HappyHorse-1.0 is in the internal testing phase, but its application programming interface will be made available in the near future, Token Hub told Yicai today. The group's innovation arm has launched an exploratory plan for a new interactive paradigm for the AI era, with video generation one part of that effort and more products to follow., it noted.

Since San Francisco-based OpenAI announced that it would pull its Sora web and app services on April 26, a number of Chinese AI video models have climbed to the top of global rankings. HappyHorse recently surpassed the Elo scores of ByteDance's Seedance 2.0, Kuaishou Technology's Kling AI, and even Google's Veo 3 Fast, according to Artificial Analysis’ Video Arena leaderboard.

Leading AI video generators are increasingly converging in their core capabilities, according to Xie Siyuan, managing director of Shanghai Yijing Capital. Competition has now shifted toward capital and resource investment, including computing power, data, and the ability to keep iterating, Xie said to Yicai.

Platform operators with vast amounts of video data and computing power advantages, including TikTok owner ByteDance and main rival Kuaishou, still retain a relative edge, Xie noted.

HappyHorse generates synchronized audio-visual videos based on text prompts. The previous efforts of Hangzhou-based Alibaba in the AI video generator arena had mainly centered on its Wan model family. 

Competition in the AI video model space has shifted from "who can generate a one-minute video" to "who can do it at the lowest cost, with higher efficiency, and closest to reality." 

Copyright data remains a major barrier for leading companies versus startups, Xie said, adding that top platforms have a richer content ecosystem and larger historical data pools, giving them advantages in data compliance and usability, while startups must both source high-quality video data and avoid copyright risks, making the task far more difficult.

The next phase to watch, Xie said, is real-time interactive video capability, where users can make live modifications and instant adjustments during generation, he said. That will push video creation from offline rendering toward real-time production and editing, with interaction becoming much closer to natural human expression, Xie pointed out.

After the announcement of Sora’s shutdown, Google’s Gemini 3 fell into a slower update cycle, while ByteDance’s AI creation platform Dreamina launched Seedance 2.0, which briefly became the industry’s focus of attention.

With improved model performance and Sora's exit, Dreamina is seeking greater pricing power in the AI video market, while Google and Kuaishou have opened to cut prices.

Editor: Martin Kadiev