Most of what we write about is text and code. This one is different. HappyHorse is a video model, and it's worth your attention for three reasons: speed, native audio, and price.
Alibaba's ATH group built HappyHorse and started a quiet rollout in late April. It generates video from a text prompt or an image, and it edits existing clips, up to fifteen seconds of multi-shot footage. What sets it apart from the Sora-and-Seedance crowd is that the audio isn't bolted on afterward. The model encodes text, image, video and audio in one stream, so speech, sound effects and picture come out together, with lip movement matched across seven languages including English, Chinese, Japanese and Korean.
Why people noticed
Two numbers did most of the talking. On a single H100, HappyHorse renders five seconds of 1080p video in about 38 seconds. That's two to three times faster than the mainstream models, at roughly 40 percent of the compute. And on the Artificial Analysis video leaderboard it went straight to the top with a score of 1300.
Price follows from the speed. The list rate is 0.9 yuan per second at 720p and 1.6 yuan at 1080p, and with a membership and a limited-time discount, 720p drops to around 0.44 yuan per second. For a model that beats the field on quality, that's an unusual place to be on the cost curve.
What you'd use it for
Three jobs, mainly. Text-to-video for generating a clip from a description. Image-to-video for animating a still. And editing for changing an existing clip with a prompt. The fifteen-second multi-shot support and the built-in audio make it a fit for short narrative content, ads, and the kind of AI short-drama work that's been driving a lot of the recent demand.
Coming to Turiloop
We're bringing HappyHorse to Turiloop so you can call it the same way you call our text and image models: one key, international card, pay-as-you-go, no Chinese phone number. If you've been blocked from Chinese video models by the usual account and payment walls, this removes them.
It isn't live on the platform yet. If video generation is on your roadmap, it's worth getting set up now, so the moment HappyHorse goes live you can point a request at it. Create a key on Turiloop and start with the text and image models today.