What Happy Horse V1.1 is good at — and what it's not

Happy Horse V1.1 is Alibaba's text-to-video model — the top-ranked successor to the original Happy Horse that climbed to #1 on the Artificial Analysis Video Arena blind-test leaderboard. It takes a prompt plus up to 9 reference images and generates a clip up to 1080p. The reference support lets you shape the look and keep subjects consistent without relying on text alone, and the long prompt allowance (up to 2,500 characters) makes it suited to detailed, precisely-described scenes.

The headline upgrade over 1.0 is framing: V1.1 outputs nine aspect ratios, from 21:9 ultrawide for cinematic sequences down to 9:21 for tall vertical, with 16:9, 1:1, 5:4, 4:3, 3:4 and 4:5 in between — up from five in the original. Paired with clip length of 3 to 15 seconds (one of the longer single-shot windows in the lineup) and larger 20 MB reference images, a full action or beat fits one frame format and one pass.

Where it sits: it's a strong general option among ZOOOP's video models. For a different balance, Kling V3 is a general flagship, Seedance V2.0 leads on motion physics, Wan V2.7 is the Wan flagship, and Grok Imagine is the fast, stylized pick; Pika V2.2 is the cheapest drafts. Happy Horse V1.1's sweet spot is reference-guided, longer-form text-to-video with the widest framing range, up to 1080p.

A reasonable mental model: default to Happy Horse V1.1 when you want reference-guided video with a wide choice of aspect ratios and a longer clip window at 1080p, and switch to Seedance V2.0 or Veo 3.1 when motion realism or cinematic quality is the priority.