Writing Prompts That Work for Video

The Core Vocabulary of Video Prompting

Every reliable video prompt uses language from two domains: cinematography (how the camera moves) and action description (what the subject does). Both matter equally. Most beginners write good action descriptions but forget the camera entirely — which is why their clips often feel flat and uncinematic.

Essential camera vocabulary to use in your prompts:

• Static shot — Camera doesn't move. Useful when you want the scene to speak for itself.
• Slow zoom in / push in — Camera moves toward the subject. Creates intimacy and tension.
• Pull back / dolly out — Camera moves away from subject. Reveals context.
• Pan left / pan right — Camera rotates horizontally. Good for following movement.
• Tilt up / tilt down — Camera rotates vertically.
• Tracking shot — Camera moves alongside a moving subject.
• Crane shot — Camera moves upward while pulling back, creating a dramatic reveal.
• Handheld — Slight natural camera shake, creating a documentary feel.

Describing Motion in the Scene

Subject motion is as important as camera motion. Use specific, visual action verbs:

• Good: "walks slowly," "gestures while talking," "waves gently in the wind," "pours into the glass"
• Weak: "moves," "does something," "is active"

Slow motion is your friend. AI video handles slow motion remarkably well because the motion is smoother and any imperfections are less visible. Add "slow motion" or "120fps" to prompts with fast action.

Time-lapse and timelapse also work well, particularly for clouds, city traffic, and natural environments.

Style and Mood Modifiers That Work

Certain style keywords consistently improve output quality across platforms:

• Cinematic — Increases production quality, adds letterbox feeling
• Depth of field — Produces background blur, more professional look
• Golden hour / blue hour — Specific, reliable lighting references
• Documentary — Natural light, observational camera behavior
• 4K, hyperrealistic — Quality modifiers that tend to improve output

A barista carefully pours steamed milk into an espresso in slow motion, overhead shot directly above the cup, dark moody lighting from a single tungsten lamp, cinematic depth of field, rich warm color grade, 4K.

Platform Differences to Know

The same prompt produces different results on different platforms. Key differences:

• Kling responds well to explicit camera direction language ("AI Director: dolly shot moving right"). Use its dedicated camera control settings when available rather than embedding camera instructions in text.
• Runway is optimized for motion prompts when used with image-to-video. Keep text prompts focused on motion and mood.
• Veo 3 handles complex multi-character scenes and physical interactions better than others. Describe physics when it matters: "water splashes realistically," "fabric drapes naturally."
• Pika treats style keywords like "Pikaffects: melt" or "Pikadditions: fireworks" as special commands. Check the platform's specific keyword documentation.

Try This Now

Rewrite this weak prompt into a strong one before generating:

A person walks outside near some trees in nature.

Your rewrite should add: who the person is, what they're doing specifically, what the camera does, what the light is like, and the mood. Then generate both versions and compare. The difference should be dramatic.