(Yicai) April 10 — HappyHorse-1.0, the artificial intelligence video generator that went viral after taking first place in multiple global rankings, was developed by Alibaba Token Hub, a new business group formed by the Chinese tech giant last month to make AI a core strategic priority.
HappyHorse-1.0 is in the internal testing phase, but its application programming interface will be available in the near future, Token Hub said today. The group’s innovative business unit has launched plans to explore new interactive paradigms in the AI era, and said video models are also part of this direction, with more products to be rolled out in the future.
Since OpenAI announced on April 26th that it would be discontinuing the Sora web and app experience, multiple Chinese AI video models have topped various global rankings. However, according to the Artificial Analysis Video Arena listing, 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.
The major AI video generators are gradually converging in their basic functions, Xie Siyuan, managing director of Shanghai Yijing Capital, told Yicai. The crux of the competition has shifted to investment in capital and resources, such as computing power, data, and continuous iteration capabilities, Xie said.
Xie pointed out that platform-based companies with vast video data and computing power advantages, such as TikTok owner ByteDance and short video platform Kuaishou, still have a relative advantage.
HappyHorse-1.0 generates audio-visual synchronized videos based on text prompts. Hangzhou-based Alibaba Group Holding’s previous efforts in the AI video generator field have mainly focused on the Wan model series.
The competition in the AI video modeling space has shifted from “who can produce a one-minute video” to “who can produce a video that is least costly, more efficient, and most similar to reality.”
Xie said that copyright data has become a key barrier for large enterprises to startups, adding that while top platforms have more complete content ecosystems and rich historical data accumulation, offering advantages in terms of data compliance and ease of use, startups face additional challenges as they need to obtain high-quality video data while avoiding copyright risks.
A notable next step will be real-time interactive video capabilities, which will include a generation process with stronger interactivity, allowing for real-time changes and instantaneous adjustments, Xie emphasized. Video generation will gradually move from offline rendering to real-time creation and editing, and interaction methods will move closer to natural human expression, Xie added.
While Google’s Gemini 3 experienced a slow update cadence after Sora’s closure announcement, ByteDance’s AI creation platform Dreamina released Seedance 2.0 and briefly captured the industry’s attention.
Improving model performance and the withdrawal of Sora also prompted Dreamina to seek pricing power in the AI video space. In contrast, Google and Kuaishou have chosen to lower their prices.
Editor: Martin Kadyev
