The difference between a mediocre AI video and a stunning one almost always comes down to the prompt. After generating thousands of videos across Kling 3.0, Google Veo, and Sora, we have compiled the prompts that consistently produce the best results.
Each prompt below has been tested and refined. Copy them directly, or use them as templates for your own creative work.
Cinematic landscape prompts
Prompt 1: "Sweeping aerial shot over a misty mountain range at golden hour. Camera slowly pushes forward revealing a hidden valley with a winding river. Volumetric fog, warm sunlight filtering through clouds, 4K cinematic, shallow depth of field at the edges."
Prompt 2: "Low-angle tracking shot through a dense bamboo forest. Camera glides smoothly at waist height, bamboo stalks passing on both sides. Dappled sunlight, gentle breeze causing slight movement, green and gold color palette, anamorphic lens flare."
Prompt 3: "Timelapse of desert sand dunes from dawn to dusk. Shadows stretching and shrinking across orange sand. Stars appearing as the sky transitions from blue to deep purple. 8K, smooth time progression, no cuts."
Why these work: they specify camera movement, lighting conditions, and color palette. The key phrase "camera slowly pushes forward" tells the AI exactly how to move, rather than leaving it to interpret "cinematic" on its own.
Product showcase prompts
Prompt 4: "Elegant product orbit shot of a matte black wireless headphone on a reflective dark surface. Camera rotates 180 degrees at table height. Studio lighting with soft key light from the left, rim light from behind. Minimal, luxury aesthetic, 4K."
Prompt 5: "Close-up of a premium wristwatch with the second hand moving smoothly. Camera slowly pulls back to reveal the full watch face. Macro photography style, dramatic side lighting, bokeh background, metallic reflections."
Prompt 6: "A glass perfume bottle sits on white marble. Camera dollies in slowly from medium shot to extreme close-up of the liquid inside. Light refracts through the glass creating rainbow caustics on the surface. Clean, luxurious, still life photography."
Tip: Product videos perform best on Kling 3.0 due to its strong object consistency and 4K output. Use the start frame feature to control the exact initial composition.
Human portrait and lifestyle prompts
Prompt 7: "Profile shot of a person in a dark hoodie standing on a rooftop at night. City lights blurred in the background. Gentle wind moving fabric. Face hidden in shadow, only silhouette visible. Moody, cinematic, blue and orange color grading, anamorphic bokeh."
Prompt 8: "A barista pours latte art in slow motion. Camera positioned at counter level, tight framing on the cup. Steam rising, warm overhead lighting, shallow depth of field. Coffee shop ambient atmosphere, 4K."
Prompt 9: "Street photographer walks through rain-soaked Tokyo alley at night. Camera follows from behind at a distance. Neon signs reflect in puddles, umbrella in hand, wet asphalt, cinematic color grading with teal and orange tones."
Note: When generating human subjects, keep faces partially hidden or in profile for the most consistent results. AI models handle silhouettes and atmospheric portraits better than direct facial close-ups.
Nature and wildlife prompts
Prompt 10: "Underwater shot of sunlight filtering through crystal clear ocean water. Rays of light dance on a sandy bottom. Small fish swim through the light beams. Camera slowly tilts upward toward the surface. Blue and turquoise color palette, peaceful movement."
Prompt 11: "Extreme close-up of a monarch butterfly landing on a wildflower. Wings opening and closing in slow motion. Macro photography with creamy bokeh background. Morning dew on petals, golden hour lighting."
Prompt 12: "Northern lights dancing over a still lake in Iceland. Time-lapse movement of green and purple aurora against a starry sky. Perfect reflection in the water. Wide-angle lens, no camera movement, 15 seconds."
Abstract and artistic prompts
Prompt 13: "Liquid mercury flows across a dark surface in slow motion, splitting and merging. Highly reflective surface catching colored studio lights. Close-up macro shot, satisfying fluid dynamics. Silver and holographic color palette."
Prompt 14: "Particles of golden dust slowly swirl in a beam of light against a black background. Camera pushes through the particle cloud. Ethereal, dreamlike atmosphere, shallow depth of field, particles glowing as they catch the light."
Prompt 15: "A single drop of paint falls into milk in extreme slow motion. Colors expand outward in organic fractal patterns. Top-down camera angle, studio macro photography, vivid colors against white, high contrast."
Architecture and urban prompts
Prompt 16: "Smooth dolly shot through a minimalist modern architecture space. White concrete walls, large windows casting geometric shadow patterns on the floor. A single person walks through the space in the distance. Quiet, contemplative atmosphere."
Prompt 17: "Low-angle fisheye shot of a futuristic skyscraper at blue hour. Glass facade reflecting the twilight sky. Camera slowly tilts upward from ground level. Ultra-wide 12mm lens distortion, dramatic perspective, wet asphalt reflections."
Prompt engineering tips
Structure every prompt in this order: subject first, then action, then camera, then style. The AI processes prompts front-to-back, giving more weight to earlier elements. If the camera movement matters most, put it before the style description.
Use specific camera terminology: "dolly in," "tracking shot," "crane up," "orbit." Avoid vague terms like "dynamic" or "cinematic" without specifying what that means. Include lens references (fisheye, 85mm portrait, anamorphic) for more predictable results.
Always specify duration expectations and resolution. Add "smooth, gentle movement" for slow-motion effects. Use "no cuts" for single continuous shots. Mention "8K" or "4K" for maximum detail even if the model's actual output resolution is lower — it affects the level of detail the AI targets.
Test these prompts on Tona.AI where you can quickly switch between Kling 3.0, Google Veo, and other models to see which produces the best result for each specific prompt style.