HappyHorse 1.1

This architecture and performance information is based on HappyHorse 1.0. The HappyHorse API is a generative video service from Alibaba's ATH-AI division (originating from Taobao & Tmall's Future Lifestyle Lab). It exposes a single-stream multimodal Transformer that generates 1080p video and synchronized native audio in one forward pass.

The HappyHorse API accepts these core parameters: model (happyhorse-1.1), prompt (text description, ≤2500 chars), resolution (720P or 1080P, default 1080P), duration (clip length 3–15s, default 5s), size (16:9 / 9:16 / 1:1 / 4:3 / 3:4, for text-to-video and reference-image generation), plus optional seed and watermark. Pass first_frame_image for image-to-video, where the first frame determines the aspect ratio, or image_urls with 1–9 images for reference-image generation. This model does not accept video_url or audio_setting.

This architecture and performance information is based on HappyHorse 1.0. Most video models generate visuals first, then add audio in a separate step. HappyHorse generates visuals and audio jointly in the same Transformer — yielding sub-pixel lip-sync, naturally aligned ambient sound, and far more believable dialogue scenes.

This architecture and performance information is based on HappyHorse 1.0. HappyHorse supports seven languages: English, Mandarin Chinese, Cantonese, Japanese, Korean, German, and French. Mouth shapes align at sub-pixel precision based on the chosen audio language.

This architecture and performance information is based on HappyHorse 1.0. On a single NVIDIA H100 GPU, HappyHorse generates a native 1080p clip in roughly 38 seconds. The 8-step DMD-2 sampler skips classifier-free guidance, making it dramatically faster than diffusion video models that require 30+ steps.

This architecture and performance information is based on HappyHorse 1.0. On Artificial Analysis blind evaluations, HappyHorse ranks #1 on both text-to-video (1333 Elo) and image-to-video (1392 Elo). Its key differentiator vs. Sora 2 and Veo 3.1 is native audio co-generation; vs. Kling, it offers stronger dialogue and lip-sync.

The two models can be called with the same API key. happyhorse-1.0 supports text-to-video, first-frame image-to-video, reference images, and video editing with a source video and audio settings. happyhorse-1.1 covers text-to-video, image-to-video, and reference images but does not support video editing. This page shows the per-second prices for HappyHorse 1.1.

Check the pricing section on this page for current per-second rates. Top up credits on APIMart to start generating videos with HappyHorse.

Send a POST request to /v1/videos/generations with your APIMart API key, model name happyhorse-1.1, and your prompt. See the HappyHorse documentation for full integration examples.

Sign up for APIMart, open the API Keys page in your dashboard and create a key. The same key calls the Alibaba HappyHorse 1.1 API and every other model on the platform, with pay-as-you-go billing and no subscription.

Yes. The HappyHorse 1.1 API is billed per second of generated video. The pricing table on this page lists every specification and shows the Gold member price, 20% off. Platinum and Diamond members save even more, up to 28%. You only pay for successful requests.

You can reach us via the live chat in the bottom-right corner, email us at [email protected], or join our Discord community. Our team will get back to you as soon as possible.