Use this guide to plan how much action to put in a single video generation, and how to handle longer stories.
Use one primary action and one camera move per video clip.
Most AI video clips are 4-10 seconds long. A 5-second clip cannot contain a chase, a fight, and a sunset — do not cram a full story into one video prompt.
For longer stories: generate one shot at a time with separate video prompts, then assemble the clips in a video editor.
If the video model supports multiple shots in one generation, spell the cuts out explicitly in the prompt: "cut to...", "then a wide shot of...", "followed by...".
Longer single generations: on models that support extended clips (such as Seedance 2.5), plan the action as timed beats — for example 0–6s establish, 6–12s trigger, 12–18s reaction, 18–25s exchange, 25–30s resolution. See Seedance 2.5: advanced reference & editing controls.