How to prompt Happy Horse 1.1: animate a still with or without a prompt, and describe motion only. A reusable template, three example prompts and mistakes.
Happy Horse 1.1 turns a still into a moving shot, and its prompt is optional. You can hand it a first frame and nothing else, or add a short line that steers the motion. This guide covers how Happy Horse 1.1 reads its inputs, a motion-only template, three examples (one with no prompt at all) and the mistakes to avoid.
Happy Horse 1.1 needs a first frame. That still sets the subject, composition, light and style, and it sets the frame shape too, since the output follows the first frame. There is no last frame and no reference input.
The prompt is optional. Leave it empty and the model animates the still on its own. Write one and it steers what moves. Either way, there is nothing to gain from describing the picture again, because the frame already holds it.
Clips run from 3 to 15 seconds at 720p or 1080p. Happy Horse 1.1 does not generate sound, so the prompt is about picture and motion only.
To change an existing clip rather than animate a still, use Happy Horse Video Edit. It edits a clip by prompt with up to 5 reference images, with output up to 15 seconds.
The [element in the frame] [small motion], the [main subject] [main action], [background or secondary motion]. [Optional: camera move or speed word].
Name things that are already in the still with "the", give each one a verb, and add a speed word ("slowly", "gently") where pace matters. Two or three moving elements are usually enough for a short clip. Leave out style, color and setting, because the frame has them.
Made in Cinemagiq. The exact prompt is under each one.
PromptNo prompt: the model animated the still on its own.
First frame made with GPT Image 2.5 Flare in Cinemagiq.
PromptThe fairy lights twinkle, the owl opens one eye and ruffles its feathers, the pages of an open book flutter.
First frame made with Nano Banana 2 in Cinemagiq.
PromptThe whale glides slowly through the clouds, its tail sweeping, the children wave from the rooftops.
First frame made with Grok Imagine 2.0 in Cinemagiq.
Example 1 has an empty prompt on purpose. Its first frame, a watercolor illustration made with GPT Image 2.5 Flare in Cinemagiq, goes in alone, and Happy Horse 1.1 animates the still on its own. It is the baseline: run a take like this first to see what the model does with a frame before you decide how much to steer it.
Example 2 starts from a storybook still made with Nano Banana 2 in Cinemagiq. It moves three things at three scales: a background shimmer ("The fairy lights twinkle"), a character beat ("the owl opens one eye and ruffles its feathers") and a prop ("the pages of an open book flutter"). Each element is named with "the", because it is already in the frame.
Example 3 animates an illustration made with Grok Imagine 2.0 in Cinemagiq. It gives the main subject a speed word, "The whale glides slowly through the clouds", adds a detail motion ("its tail sweeping") and gives the background its own action: "the children wave from the rooftops". One slow, large motion leads, and a small human action fills the rest of the frame.
For camera moves and pacing words to steer the motion, see the prompt engineering cheat sheet.
No. Happy Horse 1.1 needs a first frame, but the prompt is optional. With an empty prompt it animates the still on its own. Add a short motion prompt when you want to steer what moves, in what order and how fast.
No. Happy Horse 1.1 produces video without sound, so keep the prompt about picture and motion only. Plan the audio separately, for example with ElevenLabs Sound Effects, ElevenLabs Music or ElevenLabs Voiceover on the Generation Studio's Audio tab.
Clips run from 3 to 15 seconds at 720p or 1080p. The output follows your first frame, so compose the still at the aspect ratio you want before you animate it, and draft at 720p before you render the keeper at 1080p.
Yes, through Happy Horse Video Edit. It changes an existing clip by prompt and can use up to 5 reference images, with output up to 15 seconds. Happy Horse 1.1 itself always starts from a still image used as the first frame.
Run Happy Horse 1.1 and the other leading models in one AI-native workspace.