Breaking

Alibaba Releases HappyHorse 1.1 Video Model With Better Motion, Consistency

June 22, 2026 at 13:00 EDT

Media Generation
Foundation Models

Alibaba on June 22, 2026 released its AI video generation model HappyHorse 1.1, an upgrade that improves visual and motion expressiveness, character and text stability, and prompt adherence. It is available via the Alibaba Cloud Model Studio API and across several third-party platforms.

Jun 22 · Alibaba Group

HappyHorse 1.1 — Alibaba sharpens its AI video model

An upgrade to the model that topped blind video leaderboards: smoother motion, stronger character consistency, tighter prompt adherence, and a unified Transformer that generates video and audio in a single pass.

15B

Parameters — video + audio generated together

Reference images in R2V mode

3–15s

Clip length, 720p / 1080p, flexible aspect ratios

Version 1.0 led the blind Video Arena

Elo scores (without audio) — topped both categories vs. Seedance 2.0 and Kling.

1389

Text-to-Video

1416

Image-to-Video

New Elo scores for 1.1 not yet published at release.

What 1.1 improves

Motion
Smoother, more consistent movement

Characters
Stable across multiple reference images

Prompts
Better adherence to long instructions

Audio
Refined multilingual lip-sync

Praised

Smooth motion, strong subject preservation, cinematic quality, and more natural lip-sync — well suited to short dramas, ads, and game CG trailers.

Caveats

Motion is more natural but still falls short of Seedance in some scenes, and updated 1.1 Elo scores have yet to be published.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.