How to Write AI Video Prompts for Motion Design
· 7 min read · By the Eizoki team
A good AI video prompt reads like a shot description from a director, not a list of keywords. It says what is in the frame, how it moves, how the camera moves, what the light and materials look like, and how fast it all happens. This guide gives you a structure that works across AI video tools, a few example prompts, and a method for adapting a prompt you found to your own project.
One prompt, one shot
The most common mistake is asking for a whole video in one prompt. Models generate shots, and each shot handles one idea well: one subject, one action, one camera move. If you need a sequence, write a prompt per shot and cut them together in an editor. Short, specific shots are easier to regenerate when one of them misses.
The six parts of a motion design prompt
- Subject. What is in frame. Be concrete: “a translucent glass sphere” beats “a cool object”.
- Motion. What the subject does. Use verbs with speed and direction: “slowly unfolds outward”, “ripples from the centre”.
- Camera. One move, named plainly: slow dolly in, orbit left, static locked-off shot, top-down, macro close-up.
- Light and environment. Where the light comes from and what surrounds the subject: soft studio light from above, dark seamless background, volumetric haze.
- Style and material. The surface and finish: brushed aluminium, frosted glass, matte clay, film grain, a specific palette.
- Pacing. Slow and calm, or snappy. Mention seamless looping if you need a loop.
Some tools also accept a negative prompt: a list of things to avoid, such as text, watermarks, extra objects or camera shake. When a creator shared one, Eizoki shows it under the prompt.
Example prompts
These are written to show the structure. Adjust the wording to the tool you use, since each model responds to phrasing a little differently.
A field of thin chrome ribbons slowly twisting in unison, ripples travelling from left to right. Static camera, slight macro depth of field. Soft studio light from above, deep navy seamless background. Polished metal with subtle iridescence. Calm, slow, seamless loop.
A matte white wireless speaker on a stone plinth, fine dust drifting in the air. Slow dolly in from a wide shot to a close-up of the fabric grille. Warm low sunlight raking from the right, long soft shadows, dark warm background. Photoreal, shallow depth of field. Unhurried, premium pacing.
Soft gradient fog in coral and violet drifting slowly across the frame, gentle grain. Static camera. Empty centre of the frame for text to be added later. Minimal, calm, seamless loop.
Notice the last one leaves space for type instead of asking the model to write it. Generated text is still unreliable; most creators add it in an editor afterwards.
Lock the look with an image first
If you care about a precise style, generate or design a still frame first and animate it with an image-to-video model. The still fixes the composition, palette and materials, so your prompt only has to describe the motion and the camera. This is also the most reliable way to keep several shots consistent.
Iterate one variable at a time
- Keep a style block, the light, material and palette sentences, identical across every shot in a piece.
- When a take misses, change only one part: the camera move, or the speed, or the subject's action. Otherwise you cannot tell which change helped.
- Generate several takes per prompt. Selection is part of the craft.
Common mistakes
- Several actions or camera moves in one prompt.
- Vague adjectives (“epic”, “amazing”) instead of light, material and motion.
- Contradictions, such as a static camera that also orbits.
- Asking the model to render your logo or exact interface.
Adapting a prompt you found
A prompt that produced a great video was written for someone else's project. To reuse it well, separate its structure (camera, pacing, light, composition) from its specifics (product, brand colours, names, copy). Keep the structure and replace every specific with your own.
On Eizoki, the Copy prompt button does this framing for you: it copies the original prompt as a template together with the reference video, and tells the AI to ask which project the video is for and to replace every project-specific detail. Prompts labelled Original came from the creator; Reconstructed prompts were rebuilt from their notes, so treat them as a close starting point rather than the exact recipe.
Browse prompts by type in AI motion design, cinematic AI films and 3D experiments, or read what AI motion design is for the bigger picture.
