Motion that has to look real
Falling, pouring, collisions, cloth, hair, water — the shots where an audience notices immediately when the physics are wrong.
HappyHorse 1.1 generates video where things move as though they have mass. Start from a written scene, animate a first frame, or hand it up to nine reference images and compose a cast — at 720p or 1080p.
Creative workspace
HappyHorse 1.1 is a video generation model built for physical realism: it generates clips where weight, momentum, and contact behave the way they do in the world, from a text prompt, a first-frame image, or up to nine reference images.
How the people who built the model describe it:
“HappyHorse 1.1 generates physically realistic, fluid-motion video from text, a first-frame image, or 1~9 reference images.”
These are the jobs people actually bring to HappyHorse 1.1, and the settings each one needs.
Falling, pouring, collisions, cloth, hair, water — the shots where an audience notices immediately when the physics are wrong.
Give it up to nine reference images and address them in the prompt as character1, character2, and so on, so the same faces turn up across a series.
Hand it a still you have already composed and let the motion grow out of that frame rather than starting the look from scratch.
HappyHorse 1.1
Three ways in, one model behind them. What changes between them is how much of the look you decide up front.
01
A clip can be perfectly smooth and still look wrong, because the weight is off. HappyHorse is tuned for the other thing: momentum that carries, contact that lands, motion that settles the way mass settles.
02
Describe the scene, the motion, and the mood in plain language. Nothing to upload — useful when the shot only exists in your head and you want to see it before committing to art direction.
03
Give it a first frame and the composition, the subject, and the lighting all survive into the clip. You are directing what happens next, not renegotiating what the shot looks like.
04
Up to nine ordered reference images, called out in the prompt as character1, character2, and so on. It is how you keep several specific subjects recognisable in the same shot instead of hoping the model guesses.
HappyHorse 1.1
Where HappyHorse 1.1 is the right call, and where a different model does the job better.
Input routes
Text, first frame, or 1–9 reference images
Resolution
720P and 1080P
Native audio
Not generated — pair with a model that does if you need sound
Best decision signal
The shot lives or dies on whether the motion looks physically right
Write the idea and add any source or reference material the tool needs.
Choose the model, format, resolution, and duration that match delivery.
Review the complete draft in the shared conversation flow. Generation will be enabled when the model API is connected.
Set up the brief now, then continue in the shared Studio workspace.