25 useful AI video prompt patterns

Good prompts describe what changes over time. Use these patterns as starting points and adapt them to your source image rather than copying every line into every generation.

Camera movement

1. “Slow cinematic push-in toward the subject, stable framing, natural movement.”
2. “Gentle tracking shot moving beside the subject at walking speed.”
3. “Subtle pan from left to right while the subject remains the focus.”
4. “Mostly locked camera with slight handheld micro-movement.”
5. “Slow pull-back revealing more of the environment.”

Portrait motion

1. “Natural blink and slight head turn, relaxed expression.”
2. “Small posture adjustment and subtle breathing motion.”
3. “Gentle hair movement from a light breeze, face remains stable.”
4. “Subject looks toward the camera, then slightly past it.”
5. “Soft smile develops naturally without changing facial identity.”

City scenes

1. “Wet pavement reflections shimmer subtly as the camera moves forward.”
2. “Background pedestrians remain soft and secondary to the subject.”
3. “Neon signs create realistic reflected light on nearby surfaces.”
4. “Slow nighttime traffic movement in the distant background.”
5. “Light rain with restrained motion and no exaggerated storm effects.”

Atmospheric scenes

1. “Thin fog drifts slowly through the background.”
2. “Warm sunset light changes subtly as clouds move.”
3. “Soft dust particles catch the light, understated and realistic.”
4. “Tree leaves move gently in a natural breeze.”
5. “Water surface moves slowly with realistic reflections.”

Two-person scenes

1. “Both subjects walk side by side with natural independent movement.”
2. “The subjects share a brief glance while the camera tracks slowly.”
3. “Relaxed standing pose with subtle interaction and natural expressions.”
4. “Gentle movement in clothing and hair while both faces remain stable.”
5. “Slow camera push-in as both subjects remain clearly separated and recognizable.”

Combine only what you need

A strong five-second prompt may use one camera pattern, one subject pattern and one environmental pattern. Adding ten competing actions usually gives the model less clarity, not more.

For more fundamentals, read How to Make Cinematic AI Videos From a Single Photo.