Browse documentation
Media generation
Generate video
Plan motion, timing, source continuity, format, and review criteria for generated video.
Last reviewed September 25, 2026
Video generation adds time and motion to the creative brief. A strong prompt describes not only what appears, but what changes from the opening frame to the end.
Plan the shot
Define the subject, action, environment, camera behavior, pace, and ending state. Use concrete motion such as “the camera slowly pushes toward the product while the fabric moves in a light breeze.” Avoid stacking several unrelated scenes into one short generation.
Choose the starting material
- Text-led video is useful for open exploration when no exact starting frame is required.
- Image-led video is useful when composition, product appearance, or a key frame should anchor the shot.
- Video-led transformation is useful only when the selected route accepts video input and you need to preserve or alter existing motion.
The inputs shown in the composer are the source of truth for the selected model. If a route does not accept the reference type you need, choose a compatible model rather than assuming the input will be ignored safely.
Set duration and aspect ratio deliberately
Duration affects storytelling room, generation time, and estimated cost. Use the shortest duration that can communicate the action cleanly. Choose the destination aspect ratio before generating so the camera and subject are composed for the final frame.
Write for continuity
Keep the number of subjects manageable. Describe which objects must stay fixed, how the camera moves, and which motion should be subtle. If continuity is critical across several shots, generate them as separate controlled clips and connect them with an Assembly node.
Review frame by frame
Watch for identity drift, object deformation, inconsistent labels, abrupt camera changes, unstable backgrounds, unnatural physics, and weak opening or closing frames. Listen for audio only when the chosen route produces it; do not assume every video model includes sound.
Use Assemble video clips to order finished clips, or Create a lip-sync video when dialogue should drive a visible speaker.