HappyHorse 1.1

What is HappyHorse 1.1?#

HappyHorse 1.1 is the June 23, 2026 update to Alibaba's HappyHorse video model, built by the company's Token Hub team. It outputs 720P or 1080P video from 3 to 15 seconds in three modes: text-to-video, image-to-video from a first frame, and reference-to-video from up to nine images. Audio is native. Dialogue, ambient sound, and lip-sync render in the same pass as the picture, so a talking video needs one call, not two.

The 1.1 release improves motion, keeps subjects steadier across shots, follows prompts more closely, and cleans up the waxy faces and over-sharpened textures that 1.0 could produce. In blind-vote testing it ranks second for text-to-video with audio, behind Seedance 2.0.

Key features of HappyHorse 1.1#

Reference-to-video with up to nine images

Pass 1 to 9 reference images, up to 20 MB each, and tag them in the prompt as [Image 1], [Image 2], and so on. Characters, products, and scene elements from the images carry into the finished video.

Native audio and multilingual lip-sync

Spoken lines written in the prompt come back as synced dialogue with matching mouth movement, plus ambient sound. Lip-sync works across multiple languages, and no separate audio render is needed.

Nine aspect ratios, ultrawide to ultra-vertical

Text-to-video and reference-to-video accept 16:9, 9:16, 1:1, 4:3, 3:4, 4:5, 5:4, 21:9, and 9:21. That covers cinema strips, feed squares, and full-height vertical in one model.

First-frame image-to-video

One still becomes a moving video that keeps the source composition and aspect ratio. A prompt is optional and steers motion, mood, and sound.

Best for#

Use cases#

A micro-drama studio can hold a cast steady with nine reference stills per episode, then generate new scenes from script prompts while faces and wardrobe stay on model. An e-commerce pipeline can turn a flat product photo into a 5-second 1080P demo with camera drift and room tone. A training app can render a presenter from text alone, with timed dialogue driving lip-sync and no recorded talent. A social team can render the same brief at 9:16, 1:1, and 21:9 by changing one ratio value per request.

Limitations#

Output tops out at 1080P. There is no 4K option.

Image-to-video has no ratio parameter; output follows the source image. If you need vertical output from a still, start with a vertical frame.

Image-to-video also ranks lower in blind-vote testing than the model's text-to-video results, so prompt-only generation is its stronger mode.

HappyHorse 1.1 vs HappyHorse 1.0#

The API shape is the same: three modes, 720P or 1080P, 3 to 15 seconds, seed, watermark toggle, and up to nine reference images. Two things changed at the parameter level. Aspect ratio coverage grew from five options to nine, adding 4:5, 5:4, 21:9, and 9:21. And reference-to-video prompts now tag images as [Image 1] and [Image 2] instead of 1.0's character1 and character2 syntax.

Output quality is the bigger difference: smoother motion, steadier subjects, closer prompt adherence, more natural skin texture, and tighter audio sync. Use 1.0 only if a pipeline depends on its characterN prompt format.

When to use HappyHorse 1.1#

Use it when the video needs synced sound and picture in one call, when several reference images must appear together, or when the delivery format is unusual enough to need 21:9 or 9:21. If your source is a single still and framing must change, generate the still at the target ratio first, since image-to-video follows the input frame.

API examples#

Call HappyHorse 1.1 from any language by POSTing to /v1/tasks. Full parameter docs live at docs.unifically.com/models/video/alibaba/happyhorse-1.1-video.

curl -X POST https://api.unifically.com/v1/tasks \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -d '{
    "model": "alibaba/happyhorse-1.1-video",
    "input": {
      "mode": "t2v",
      "prompt": "A golden retriever running through a field of wildflowers at sunset",
      "resolution": "1080P",
      "ratio": "16:9",
      "duration": 5
    }
  }'

Successful submission returns a task_id. Poll GET /v1/tasks/<task_id> or set a callback_url on the request to receive the finished result.