Alibaba's newest AI video generation model, HappyHorse-1.0, has stormed the leaderboard of Artificial Analysis' Video Arena, securing either first or second place across all categories.
No-Audio First, Audio Tier Ties Seedance
The evaluation reveals a two-tier competitive landscape. In the no-audio ranking, HappyHorse-1.0 leads by a wide margin. In the audio ranking, its Elo score is nearly identical to ByteDance's Dreamina Seedance 2.0, placing the two models in a dead heat.
Functionally, HappyHorse-1.0 supports four generation modes: Text to Video and Image to Video, each offered with or without native audio. These cover the most common use cases. Full API access is scheduled for April 30.
Alibaba's Video Generation Strategy
Alibaba has opted for a name that strips away technical jargon. This marks a departure from its earlier Wan series, which emphasized technical frameworks. HappyHorse is more of a brand play targeting creator communities. Video generation is among the most hotly contested AI arenas, with Google, OpenAI, Runway, ByteDance, and Kuaishou all in the game. Alibaba's HappyHorse-1.0 aims to recapture attention that recently shifted to Seedance.
In side-by-side comparisons, HappyHorse demonstrates notable strengths in motion coherence, scene consistency, and natural facial expressions. Its lead is especially clear in the no-audio scoring dimension.