(Yicai) April 10 -- HappyHorse-1.0, an artificial intelligence video generator that has gone viral after topping multiple global rankings, was developed by Alibaba Token Hub, a new business group set up by the Chinese tech giant to give AI a core strategic priority last month.

HappyHorse-1.0 is in the internal testing phase, but its application programming interface will be made available in the near future, Token Hub said today. The group's innovative business division has launched an exploration plan for a new interactive paradigm in the AI era, with the video model being part of this direction, while more products will be rolled out later, it noted.

Since OpenAI announced that it will discontinue the Sora web and app experiences on April 26, multiple Chinese AI video models have taken the lead in various global rankings. However, HappyHorse-1.0 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 a list by Artificial Analysis Video Arena.

Leading AI video generators have gradually converged in their fundamental capabilities, Xie Siyuan, managing director of Shanghai Yijing Capital, said to Yicai. The core of competition has shifted to investments in capital and resources, including computational power, data, and continuous iteration capabilities, Xie noted. 

Platform-based companies with massive video data and computational power advantages, including TikTok owner ByteDance and short-video platform Kuaishou, still hold a relative edge, Xie pointed out.

HappyHorse-1.0 generates audio-visual synchronized videos based on text prompts. The previous efforts of Hangzhou-based Alibaba Group Holding in the AI video generator area were primarily focused on its Wan model series.

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 forms a key barrier for leading companies against startups, Xie said, adding that top platforms have a more complete content ecosystem and a richer historical data accumulation, giving them advantages in data compliance and usability, while startups face increased challenges due to needing to acquire high-quality video data while also avoiding copyright risks.

The next phase worth watching is real-time interactive video capabilities, involving a generation process with stronger interactivity, allowing for real-time modifications and instant adjustments, Xie stressed. Video generation will gradually shift from offline rendering to real-time creation and editing, with interaction methods becoming closer to natural human expression, Xie added.

Google's Gemini 3 experienced a slow update rhythm after the announcement of Sora's closing, while ByteDance's AI creation platform Dreamina launched Seedance 2.0, which briefly became the focus of the industry. 

Improvement in model performance and Sora's exit also prompted Dreamina to attempt to gain pricing power in the AI video sector. In comparison, Google and Kuaishou opted to reduce prices.

Editor: Martin Kadiev