HappyHorse 1.0

A prompt-to-clip model with no reference images and no video input: you describe the shot and it renders it. That makes it a poor fit for anything that has to match existing material, and a reasonable one for original shots where the prompt is the whole specification.

Five aspect ratios including 4:3 and 3:4, which is more coverage than the Veo family offers, and a three-to-fifteen-second range that is wider than most. 720p and 1080p.

It sits in an awkward middle of the catalog: more expensive than the Seedance models it competes with on length and resolution, without their reference capacity. The case for it is that models render differently, and when a shot is not landing on the models you usually reach for, a genuinely different one is often a faster fix than another ten attempts on the same one. Treat it as an alternative rather than a default, and check the price of a take before committing to a run of them.