Happy Horse, a new cutting-edge video model, is now available on April 27th via fal’s Generated Media Cloud for developers and enterprises.
san francisco, April 27, 2026 /PRNewswire/ — On April 27, 2026, fal launched developer and enterprise access to HappyHorse-1.0. This AI video model holds #1 in Elo rankings in the artificial analytics video arena for both Text-to-Video and Image-to-Video. fal will be one of the first official API providers for this model, giving developers immediate access on launch day through its generative media platform.
Access the HappyHorse-1.0 API and Happy Horse Playground website
Developers and enterprise customers can access HappyHorse through four API endpoints on fal: image-to-video, reference-to-video, text-to-video, and video-edit. fal provides Python and JavaScript SDKs for developers to reduce manual REST integration time. The fal API endpoint actually exposes the model’s integrated multimodal functionality (lipsync and Foley sound), rather than just silent video. fal gives you the option to choose 720p or 1080p resolution and supports formats such as 16:9, 9:16, 1:1, 4:3, and 3:4, ensuring your content is suitable for a variety of social platforms. fal also guarantees full commercial rights to all output produced. fal’s lightning inference speed is optimized with the latest AI infrastructure technology, making it one of the fastest ways to generate content via the HappyHorse-1.0 API.
Official Happy Horse API endpoint
You can access the official API via fal using four API endpoints. You can obtain your API key from the API Keys page on your fal.ai dashboard.
About HappyHorse-1.0
HappyHorse-1.0, built by Alibaba Taotian Future Life Lab, part of Alibaba. token Hub Division is the first video model to reach the top of an artificial analytics leaderboard ranking based on blind human preference voting. The user compares two unlabeled clips and chooses the better one without knowing which model was created. This model scores an Elo of 1333 for Text-to-Video and 1392 for Image-to-Video (no audio), outperforming all other video generation models currently benchmarked.
HappyHorse-1.0 was led by Zhang Di, a 15-year veteran of the AI industry. Zhang Di previously served as Vice President of Kuaishou and Technical Architect of Kling AI, and returned to Alibaba in late 2025. A spokesperson for fal said:HappyHorse is known for creating 1080p videos with synchronized audio, strong lighting, and realistic, emotional, and consistent detail. ”
The core architecture of this model is an integrated 40-layer self-attention transformer that jointly generates video and audio in a single forward pass without the need for cross-attention modules or separate audio post-processing steps. The result is a synchronized audiovisual output with native lip-sync support across seven languages: English, Mandarin, Cantonese, Japanese, Korean, German, and French. The team claims that the generation time for 1080p output on a single NVIDIA H100 GPU is approximately 38 seconds.
Example of using Happy Horse
This model supports use cases ranging from product promotions and social content to multi-shot sequences with consistent character identity across the frame. Camera direction fidelity, including cues such as “slow dolly push,” “overhead crane shot,” and “light and strong winds,” are examples of key differentiators for developers building video production workflows. To start using Happy Horse, visit fal.
About Fal: fal is a generative media platform for developers and enterprises, specializing in high-performance inference and fine-tuning across images, video, audio, and 3D models. We provide low-latency APIs for cutting-edge models such as HappyHorse-1.0, Seedance 2.0, Nano Banana, and Flux, serving industries such as gaming, e-commerce, and creative production. fal is one of the first official API providers for a major model release across the generative media environment. The platform supports both self-service and enterprise customers with instant access to models, high-throughput infrastructure, and dedicated ML engineering support.
source fal
