HappyHorse-1.0The Top Ranked AI Video Model
#1 on the Artificial Analysis Video Arena in both Text-to-Video and Image-to-Video, ranked by blind human preference votes. Joint audio-video generation in a single pass. HappyHorse API is scheduled for release around end of April 2026, and performance in private beta is still being verified by fal as an official API provider.
Artificial Analysis Video Arena Rankings
Elo ratings based on blind human preference votes. Users compare two videos from the same prompt without knowing which model produced which. In video generation with the elo system, people compare two unlabeled clips and pick the better one. The winning model gains points, the loser loses some. Generated video samples posted by the benchmark providers showed Happy Horse performing well leading to #1 results for the following arenas.
Source: Artificial Analysis Video Arena, April 2026. Scores reflect early vote counts and may shift as more votes accumulate.
Why HappyHorse-1.0 Is #1
#1 in Blind Human Preference
HappyHorse-1.0 holds the top Elo rating on the Artificial Analysis Video Arena in both Text-to-Video and Image-to-Video (no audio). Rankings are based on blind preference votes from real users who do not know which model produced the output they are voting on.
Video and Sound in a Single Pass
The model reportedly generates video and audio jointly in a single forward pass using a unified 40-layer self-attention Transformer with no cross-attention modules. This architecture produces synchronized audiovisual output without separate audio post-processing.
1080p in Under 40 Seconds
The team claims approximately 38-second generation time for 1080p output on a single NVIDIA H100 GPU, and roughly 2 seconds for a 5-second clip at 256p. If verified, this would represent a significant speed advantage over current alternatives.
HappyHorse-1.0 Features
A unified audio-video transformer that generates cinematic output from text or image inputs in a single pass.
Text to Video
Describe a scene in natural language with the subject, motion, lighting, camera angle, and mood. HappyHorse-1.0 generates a 1080p video with synchronized audio in a single pass. The unified transformer processes text, video, and audio tokens together, so dialogue, music, and sound effects are generated natively rather than layered on after the fact. Supports lip-syncing and audio in seven languages: English, Mandarin, Cantonese, Japanese, Korean, German, and French.
Image to Video
Upload a reference image — a product shot, a character frame, or a concept sketch — and Happy Horse animates it with real-world physics, director control, and a multitude of camera movement styles. The image is embedded directly into the model’s instructions, keeping the subject’s identity, proportions, and visual style locked throughout the generated clip. Ideal for product promos, social content, and any workflow where you need to go from a still asset to motion.
Motion Quality and Scene Consistency
Faces, expressions, movement, fabric, and camera all have correct frame-to-frame physics, preventing glitches and telltale signs of AI-generated content. Character consistency is a strong point — identity, wardrobe, and lighting stay consistent throughout multi-shot sequences.
Prompt Adherence
What you write is what you get. Camera directions like “slow dolly push-in” or “overhead crane shot” are executed with high fidelity. The model distinguishes motion cues such as “breeze” versus “strong wind,” “walking” versus “striding.” For best results, lead with your most important instruction. This level of prompt control is one of the main reasons developers are building video generation workflows on Happy Horse.
The Team Behind Happy Horse
On April 9, 2026, Alibaba Group Holding Ltd revealed it created the “Happy Horse” video AI model that sent ripples across the AI industry, claiming ownership of a platform that hit #1 on global rankings for its debut. Happy Horse is the product of the nascent Alibaba Token Hub's innovation business unit, and still under a beta testing phase.
What is Alibaba Token Hub (ATH)?
Alibaba Token Hub is a high-level division that brings together all of Alibaba's AI expertise, from research labs to real-world software, under one roof. Led by CEO Eddie Wu, it's designed to turn advanced models like Qwen into practical, everyday tools by focusing on the “token” as the essential fuel for the modern AI economy.
Led by Zhang Di
Zhang Di is a veteran AI engineer with 15+ years in the field. He served as Director at Alibaba Group from 2010–2022, then joined Kuaishou as Vice President where he was the technical architect of Kling AI. He rejoined Alibaba in late 2025 to lead the Taotian Future Life Lab under ATH, and within months delivered Happy Horse 1.0.
Open-Source Status
While other industry players say HappyHorse-1.0 will be open source, we can confirm that HappyHorse-1.0 will be closed source. It will not be licensable or open source.
See what HappyHorse-1.0 can create
Sample outputs from the Artificial Analysis Video Arena and community-shared generations.
Common questions about HappyHorse-1.0
What is HappyHorse-1.0?
HappyHorse-1.0 is an AI video generation model that appeared on the Artificial Analysis Video Arena on April 7, 2026, immediately ranking #1 in both Text-to-Video and Image-to-Video (no audio) categories. It uses blind human preference voting where real users compare outputs without knowing which model produced them.
Who built HappyHorse-1.0?
The model was submitted pseudonymously to the Artificial Analysis leaderboard. The team's own marketing materials claim it was built by the Future Life Lab team at Taotian Group (Alibaba), led by Zhang Di, described as the former VP of Kuaishou and technical lead of Kling AI. This claim has not been independently verified.
What are the technical specs?
According to the team's own sites: 15 billion parameters, a unified 40-layer self-attention Transformer that generates video and audio jointly in a single forward pass with no cross-attention modules. Claimed inference speed is approximately 38 seconds for a 1080p clip on a single NVIDIA H100 GPU. These specs have not been independently verified.
Can I use HappyHorse-1.0 right now?
Not yet. As of April 2026, there is no public API, no downloadable model weights, and no confirmed pricing. The official HappyHorse-1.0 is expected to be available soon via fal in late April 2026.
How does the Artificial Analysis ranking work?
The Artificial Analysis Video Arena uses an Elo rating system based on blind human preference votes. Users see two videos generated from the same prompt, do not know which model produced which, and vote for the one they prefer. Rankings reflect what real people prefer under blind conditions, not self-reported benchmarks.
What languages does HappyHorse-1.0 support?
The team claims native lip-sync support across seven languages: Mandarin, Cantonese, English, Japanese, Korean, German, and French. This has not been independently tested.
Which HappyHorse-1.0 website or social media account is genuine?
While there are a lot of new websites popping up claiming to host the HappyHorse-1.0 API, none of them are official. The only official first party social media for HappyHorse-1.0 is @HappyHorseATH on X. fal is an official partner and can confirm no one has official API access yet.
Is HappyHorse-1.0 open source?
While other non-official industry players say it will be open source, the truth is that HappyHorse-1.0 will not be open source. This discrepancy comes from no official statement yet provided by Alibaba Group.
When will HappyHorse-1.0 be available on fal?
HappyHorse-1.0 is coming soon to fal. We will make it available via playground and API as soon as access is possible. Check back for updates.
Official API Access: HappyHorse-1.0 Coming to fal
fal has been among the first platforms to host each of the leading generative video model releases. We're excited that HappyHorse-1.0 API is the next. As one of the exclusive official API providers at launch (expected late April 2026), we will be enabling access for enterprises and developers from day one. Our technical team is already integrating HappyHorse-1.0's audio-visual capabilities into our generative media cloud. Enterprise and self-serve users will have immediate access to the model the moment it's available. Follow us on Twitter to hear about the official launch the minute it goes live.
Ready to transform your enterprise with AI?
Take the first step towards AI-driven innovation. Our team of ML engineers is ready to help you prototype, develop, and scale your AI solutions.