Cinematic AI video: lighting, camera and shot length

Describe where the light comes from and what it does to the subject. A large window, a small practical lamp or low evening sunlight suggests a more concrete visual setup than a stack of quality adjectives.

Keep direction and contrast consistent with the setting. If you want a silhouette, explain the backlight; if the face matters, make it readable. Generated lighting still needs inspection across the clip.

Framing describes how much of the subject and setting is visible. An angle establishes the viewpoint. Movement describes how that viewpoint changes over time. You do not need all three to change in the same shot.

Start with a static shot or a restrained push-in. A pan, tracking move and orbit have different spatial demands. Pick one rather than asking for several contradictory camera paths.

A medium shot of a fictional adult watchmaker at a workbench. Warm light from a desk lamp falls across the hands, with a cool window in the background. A slow push-in as the watchmaker places one gear. One continuous shot.

Choose a supported duration that fits the main movement. Too many actions compressed into a short clip can produce jumps or incorrect order. A longer clip is not automatically a better solution if the action itself is unclear.

For a sequence, decide where the viewer should look before planning the next shot. Use editing to create changes in viewpoint and pacing instead of expecting an overloaded generation to direct the whole scene.

Check contact with surfaces, shadows, faces and the camera path. A convincing colour grade can hide a broken action in a thumbnail. Watch the whole result at normal speed and inspect suspicious moments.

Compare attempts using the same shot brief. Record the lighting and camera instruction with accepted material so that later scenes can follow the same visual direction.