Happy Horse 1.1 — Alibaba's video model with sound

Happy Horse 1.1 is Alibaba's AI video model, released on 22 June 2026 as an upgrade to HappyHorse 1.0. It makes 3–15-second clips at up to 1080p and 24 fps with native audio, from a text prompt, a start image or reference images. On Kenerate it runs in the browser at 720p or 1080p, with no subscription required.

Alibaba makes it, through its Alibaba Token Hub (ATH) business unit. HappyHorse 1.0 entered a limited beta on 28 April 2026, and version 1.1 followed on 22 June 2026. The official site is happyhorse.com, and developers can reach the model through the API on Alibaba Cloud Model Studio. On Kenerate it runs in the video studio at /app/video.

Yes. Alibaba lists native audio generated with the picture, including dialogue, ambient sound and effects. Every clip we made came with sound: a spoken line in the portrait, a tram bell and city ambience in Lisbon, wind and a metallic ring for the samurai. Describe the sound you want in the prompt and put any spoken line in quotes.

You choose the mode, length, resolution and aspect ratio. Modes are text to video, image to video, or reference to video with up to 9 images. Clips run 3–15 seconds at 720p or 1080p, in 16:9, 9:16, 1:1, 4:3 or 3:4, and you can extend a clip. There is no end-frame control on Kenerate.

Reference to video is the headline addition: 1.1 takes up to 9 reference images and keeps those subjects consistent in the output. Alibaba also says 1.1 upgrades motion, character consistency, instruction following, visual texture and audio. Version 1.0 entered a limited beta on 28 April 2026, and 1.1 was released on 22 June 2026.

Seven: English, Mandarin, Cantonese, Japanese, Korean, German and French. The model generates the voice with the picture and matches the mouth to the spoken audio at phoneme level, so put the spoken line in quotes in the prompt and name the language if it isn't obvious.

In our tests, yes for short clips with sound. Every clip came with audio — ambience, effects and a spoken line in the portrait — and the start image's face and details held through the motion. The limits: image to video follows the start image's shape, and there is no end-frame control on Kenerate. Test a prompt at 720p before moving to 1080p.

Between 3 and 15 seconds on Kenerate, at 720p or 1080p, and you can extend a clip after it is made. Alibaba lists the same 3–15-second range for the model. Test a new prompt as a short 720p clip first, then move to 1080p or a longer length once the motion and sound look right.