1. Define one shot
Write a single sentence explaining what the viewer should see. Break a longer story into separate shots rather than asking one short generation to handle multiple locations, actions, and cuts.
2. Choose text or image input
Start with text-to-video when you are exploring a new scene. Use image-to-video when you have a permitted reference and want to start from its framing. Check the selected model’s input requirements; the two modes are not interchangeable.
3. Describe motion explicitly
Name the main subject’s action, then the camera movement, lighting, and pace. Avoid contradictory instructions such as a locked-off camera and an orbit in the same shot. Choose the supported duration and aspect ratio in the editor.
4. Review before assembling
Watch the whole clip and inspect transitions frame by frame where needed. Check subject continuity and unintended movement. Confirm whether the output includes usable audio; plan narration or music separately when it does not.
Example brief
One continuous shot of a red paper boat drifting slowly across a shallow pond. Low camera angle at water level, gentle camera push forward, small natural ripples, warm morning light, calm pace, consistent boat shape, no scene change, no text.
An illustrative starting prompt, not a guarantee of an exact output. Adjust it for the model and supported settings.
Before you publish
- Check the beginning and end of the clip, not just a thumbnail.
- Look for changing object shapes, sudden cuts, and inconsistent motion.
- Review the cost estimate for duration, resolution, and audio options before generating.