Subject
Describe the main subject clearly. Include only the details that matter visually: age range, clothing, product type, material, color, or role. Avoid adding too many competing details in a short clip.
Scene
The environment tells the model what physics and lighting to expect. “Modern kitchen at sunrise” is more useful than “nice room.” If the background is not important, keep it simple so the subject stays stable.
Shot type and camera
Use film language. Close-up, medium shot, wide shot, macro, tracking shot, slow push-in, dolly backward, handheld, orbit, and overhead shot all provide stronger direction than “make it cinematic.”
Action
One action per clip is usually enough. A subject can walk, turn, pour, lift, smile, open a laptop, or look toward camera. Too many actions increase the chance of warped motion.
Style and constraints
Style words should support the objective: documentary, premium commercial, natural handheld, editorial fashion, clean studio, or warm training video. Constraints help too: no text, stable hands, realistic motion, no warped logo, no extra people.
Example prompt
“Medium close-up of a founder explaining a new AI video workflow at a clean desk, slow push-in camera movement, warm softbox lighting, natural eye contact, realistic office background, documentary style, no text, no extra people.”
For cinematic templates, read Higgsfield prompts for cinematic AI videos.