Alibaba HappyHorse 1.1: Pricing, Capabilities, and Key Upgrades for AI Video Generation

On June 22, 2026, Alibaba released HappyHorse 1.1, a major upgrade to its AI video generation model . The update addresses common pain points from version 1.0—such as sluggish motion and inconsistent character appearance—and adds new creative features like multi-image reference-to-video and long-context instruction following. Here is a fact-checked overview of the key details, capabilities, pricing, and availability.

Release & Availability

HappyHorse 1.1 is accessible through several channels:

• HappyHorse's official website
• Alibaba Cloud Model Studio (bailian.console.aliyun.com) with full API access for enterprise users
• The Qwen app

Creative Fabrica announced it is among the first third-party platforms to offer early access to the model .

Key Upgrades from HappyHorse 1.0

Alibaba says 1.1 delivers systematic improvements across five dimensions :

• Dynamic Expression — Rebuilt motion modeling and temporal coherence to fix the sluggish movement seen in 1.0; produces smoother, more forceful action .
• Subject Consistency — Supports up to 9 reference images simultaneously to preserve characters, products, brand elements, and scenes across shots. This is critical for multi-character short dramas, livestream e-commerce, and ads .
• Instruction Following — Improved handling of both short prompts and complex narrative inputs, enabling long-context instruction following for multi-element scenes .
• Visual Quality — Upgraded skin textures for more natural-looking humans, with better overall image fidelity. The improvements address issues like "oily skin" and "over-sharpening" reported by users of version 1.0 .
• Native Audio (Beta) — Audio capabilities are now integrated directly into the generation pipeline .

Supported Generation Modes

HappyHorse 1.1 supports the same four modes as 1.0, all available via a unified API :

Mode	Input	Output
Text-to-Video (T2V)	Text prompt	Video
Image-to-Video (I2V)	1 reference image + optional prompt	Animated video
Reference-to-Video (R2V)	1–9 reference images + prompt	Subject-consistent video
Video Edit	Source video + up to 5 optional reference images + instruction	Locally/globally edited video

Resolution & Clip Length

• Duration: 3 to 15 seconds per generation
• Resolutions: 720p and 1080p, with flexible (free) aspect ratios

Standard Pricing

Pricing is per second of output (not per compute second) and is largely unchanged from HappyHorse 1.0 :

Tier	Standard price (per second)
720p	$0.14/sec (~¥0.90 RMB/sec)
1080p	$0.18/sec (~¥1.17 RMB/sec, based on Chinese reports)

Note: Some third-party resellers list 1080p at $0.28/sec , but the official Alibaba Cloud list price for 1.1 appears to be $0.14/sec for 720p and roughly $0.18/sec for 1080p, consistent with 1.0 pricing . A Chinese report notes that the 1080p price was reduced by 25% compared to version 1.0 (from ¥1.6/sec to ¥1.2/sec) .

40% Launch Discount

Alibaba Cloud Model Studio is offering a promotional 40% discount for the initial two weeks, bringing pricing down to approximately $0.084/sec for 720p and $0.108/sec for 1080p . Alibaba Cloud's own campaign page lists 720p at $0.084/sec (discounted from $0.14) and 1080p at $0.108/sec .

Caveats

• Most performance claims (generation speed, leaderboard rankings) are Alibaba's own self-reported numbers and have not been independently verified by third-party benchmarks .
• The exact $0.18/sec figure for 1080p standard pricing is inferred from Chinese pricing reports (1.17 RMB ≈ $0.18 USD); some third-party platforms charge differently .
• The native audio feature is described as in beta; third-party platforms may or may not support it .