On April 27, 2026, Alibaba's ATH Business Group officially launched a limited beta of its self-developed AI video generation model HappyHorse 1.0 on the Qwen platform. Dubbed "Happy Horse" by the industry, this video large model topped the Artificial Analysis global leaderboard with an Elo score of 1,357, surpassing ByteDance Seedance 2.0, Grok Video, and other major international competitors to become a formidable dark horse in the AI video space.
What Makes HappyHorse Reach "Global No. 1"?
HappyHorse 1.0 has delivered stunning performance across authoritative AI model evaluation leaderboards: LMArena Video Editing Board — topped with an Elo score of 1,299, leading second-place Grok Video by 42 points; Artificial Analysis Text-to-Video / Image-to-Video Board — anonymously debuted at #1 in early April, outscoring ByteDance Seedance 2.0; Official GSB Human Preference Score — reached 3x that of Wan2.7.
HappyHorse is led by Alibaba VP Zheng Bo's team and belongs to the ATH (Alibaba Token Hub) Innovation Division. This division integrates five major segments: Tongyi Lab, MaaS Platform, Qwen, Wukong, and AI Innovation, personally overseen by Alibaba Group CEO Wu Yongming.
Core Feature Highlights
1. Narrative-Level Intelligent Camera Work
HappyHorse abandons the "fixed camera" safe compositions of traditional AI video, automatically orchestrating camera language based on emotional context and environmental needs: auto-pushing to close-ups when emotions intensify; switching to wide establishing shots for scene context; delivering professional cinematographer-level camera work.
2. High-Fidelity Action and Micro-Expression Control
Character movements remain coherent and stable from start to finish, drastically reducing glitches and distortions. Supports nuanced micro-expression generation: facial muscle tension when biting a lemon, furrowed features, eye movement emotional progression; in multi-person dialogue scenes, listeners show authentic micro-expressions and muscle feedback; action transitions flow smoothly without the common AI video "moonwalk" sliding phenomenon.
3. Realistic Voiceover and Ambient Sound Narration
HappyHorse breaks through AI voiceover's "robotic reading" limitations, achieving deep audio-visual fusion: natively generated audio with no need for separate post-production voiceover; natural intonation that rises and falls naturally with context, rhythm matching visual emotion; ambient sound synchronization with details like paper rustling and distant background sounds precisely matching scene logic; multilingual lip-sync covering Mandarin, Cantonese, English, Japanese, Korean, German, and French — 7 languages total.
4. Zero-Barrier Cinematic Style Reproduction
Without complex prompts, accurately reproduces multiple classic visual aesthetics: Hong Kong film grain with cool highlights; historical drama gritty realism; American TV high-contrast cool lighting; Korean drama delicate soft-focus ambiance.
Four Ways to Use HappyHorse
Method 1: Qwen App (Recommended for Beginners)
Download and update the Qwen App to the latest version
Open the app, tap the "HappyHorse" capsule button on the home screen
Enter the creation panel, choose "Text-to-Video" or "Image-to-Video"
Enter a prompt or upload an image, set parameters, and generate
Basic parameters: Resolution up to 1080P super-resolution output; Aspect ratio 16:9 landscape, 9:16 portrait, 1:1 square freely switchable; Duration options 5s, 10s, 15s; Cost App includes official free trial quota, limited-time free use.
Method 2: Qwen Web (Professional Creation)
Open browser, visit c.qianwen.com
Log in with Alipay/Taobao account
Find the "HappyHorse" entry in the feature list
Supports more granular parameter adjustments and batch generation
Advantages: Large-screen operation with more efficient prompt editing; suitable for complex scenes and longer videos; per-use credit consumption with excellent overall value.
Method 3: Liblib Platform
Liblib, as a leading domestic AI creation community, has simultaneously launched the HappyHorse 1.0 model.
Liblib advantages: Rich model library — beyond HappyHorse, access Seedance, Veo, Wan and other mainstream video/image models simultaneously; Active community — massive user-shared prompt templates and case references lowering the learning barrier; Flexible API — supports batch generation and automated workflow integration; VIP membership perks — register via promotional links for exclusive discounts and higher quotas.
Visit Liblib VIP page to unlock HappyHorse 1.0
Method 4: RunningHub (API Access & Batch Generation)
RunningHub is a professional AI model aggregation service platform that has launched the HappyHorse 1.0 API interface.
Core advantages: Standardized API — provides RESTful API with Python/JavaScript SDK support; Batch concurrency — supports high-concurrency requests and async task queues for large-scale video generation; Transparent billing — pay-per-use model with no hidden fees and enterprise-grade billing management; Model aggregation — beyond HappyHorse, integrates Seedance, Veo, Wan and other mainstream AI video/image models for one-stop access; Referral benefits — register via promotional links for new user discounts and bonus credits.
Visit RunningHub — new user discount on registration
Recommended Use Cases
HappyHorse 1.0 focuses its core capabilities on naturalness of motion, sound, and spatial relationships, fundamentally transforming AI video's "lottery-like" unpredictability.
Short video content creation: Self-media bloggers and new media operators can rapidly produce high-quality video assets, reducing filming costs
Short drama/film pre-visualization: Directors, screenwriters, and producers can use it for storyboard pre-vis and concept short films, saving pre-production communication time
E-commerce marketing videos: Brands and e-commerce operators can efficiently generate product promos and merchandise showcase videos
AIGC freelance work: Freelancers and AIGC service providers can take on AI video customization projects, increasing service premium capability
Personal creative expression: Students and hobbyists can achieve cinematic visual dreams with zero barrier
Usage Tips and Notes
Prompt writing tips: Make scene descriptions specific (e.g., "girl walking through a flower field, skirt flowing in the wind"); use camera language guidance (e.g., "push in from wide shot to face close-up"); add style keywords (e.g., "Hong Kong film grain," "Korean drama soft-focus ambiance").
Common notes: During beta, some users may temporarily not see the entry point; if issues arise, update the app or contact official support; single generation maximum length is 15 seconds; for longer videos, use segmented generation + post-production stitching; multilingual lip-sync performs excellently, but complex dialect scenarios still recommend manual review.
"Tianma Xingkong" AIGC Video Challenge
Qwen platform has simultaneously launched the "Tianma Xingkong" (Limitless Imagination) AIGC Video Challenge:
Four creative tracks: Covering narrative short films, creative ads, music videos, and experimental visuals
200,000 yuan prize pool: Outstanding works can win generous cash prizes and official recognition
How to participate: Search "Tianma Xingkong Challenge" in the Qwen app to register
HappyHorse vs Major Competitors
HappyHorse 1.0: Elo 1,357 (top rank); narrative-level auto-orchestrated intelligent camera work; 7-language full-coverage multilingual lip-sync; natively supports ambient sound generation as part of narration; Qwen app limited-time free trial; Alibaba Cloud native deployment with ultra-fast response.
Seedance 2.0: Rank #4; partially supports intelligent camera work; 3-language multilingual lip-sync; requires post-production ambient sound; paid subscription; ByteDance cloud services.
Grok Video: Rank #2; basic fixed-camera intelligent camera work; 2-language multilingual lip-sync; no ambient sound generation; X platform membership; overseas servers.
Future Outlook
API opens April 30: Developers will be able to call HappyHorse via API and integrate it into their own applications and workflows
Continuous optimization: The team is conducting final optimization sprints; the official release will introduce more advanced features
Commercial pricing TBD: Official pricing has not yet been announced; tiered subscription plans are expected after the beta period ends
Creators are encouraged to take advantage of the current free trial window to fully explore HappyHorse's creative boundaries.
This article is based on official HappyHorse announcements and media hands-on testing. Model features may change with version updates; please refer to the latest official release.
