HappyHorse-1.0 AI Video Generator Tops Open-Source Rankings

The open-source AI model HappyHorse-1.0 has solidified its position as a leading video generator, outperforming closed-source solutions including ByteDance Seedance 2.0 in blind user preference tests and topping the Artificial Analysis Video Arena leaderboard.

In the Text-to-Video (no audio) category, HappyHorse-1.0 achieved 1333–1357 Elo points, nearly 60 points ahead of the previously dominant Seedance 2.0. The model also set a new record in the Image-to-Video category with 1391–1406 Elo and secured second place in the audio-inclusive track.

Developed by an independent research team formerly of Alibaba's Taotian Group Future Life Laboratory, led by Zhang Di (ex-Kuaishou VP and Kling AI technical architect), HappyHorse-1.0 is fully open-source with a commercial license. It utilizes a 15-billion-parameter unified Transformer architecture for native audio-video synchronization.

The model offers 1080p cinematic output in 38 seconds on a single H100 GPU. Key components, including model weights, distilled versions, a super-resolution module, and inference code, have been released on GitHub, enabling broad accessibility and commercial use.