Veo 3 Prompts, Updated for Veo 3.1: 15+ Ready-to-Use Examples and How to Write Them

Google’s Veo changed the conversation around AI video because it understands cinematic language. Lighting descriptions change the output. Camera instructions get followed. And since Veo 3, the model generates sound with the picture, so a prompt can include dialogue, footsteps and room tone.

The catch is that Veo only performs as well as the prompt you give it. A vague sentence gets a vague, generic clip. A structured prompt gets footage that looks planned.

This guide is updated for Veo 3.1, the current version. Everything that worked for Veo 3 prompts still works, and 3.1 adds reference images, first and last frame control, clip extension and vertical video. Below you will find the prompt structure Veo responds to, how to use the 3.1 features in your prompts, 15+ copy-and-paste examples, and the mistakes that waste generations.

What Veo 3.1 Is and What Changed From Veo 3

Veo 3.1 is Google DeepMind’s video model, released in October 2025. It is still the newest model in the Veo line as of September 2026. Here is what matters for prompting:

Native audio. Dialogue, sound effects, ambience and music are generated with the video. Put spoken lines in quotes and Veo lip-syncs them.
Ingredients to Video. Upload up to three reference images of a person, character or product, and Veo keeps their appearance in the clip.
Frames to Video. Give a first frame and a last frame, and Veo generates the motion between them.
Extend. Continue a clip from its last second. In the Gemini API each extension adds 7 seconds, up to 20 times, for up to 148 seconds total at 720p. Google Flow advertises videos of a minute or more.
Clip length and format. A single generation is 4, 6 or 8 seconds, in 16:9 or 9:16. 1080p and 4K are available on 8-second clips.
Three speeds. Veo 3.1, Veo 3.1 Fast and Veo 3.1 Lite. In the Gemini API they cost $0.40, $0.10 (720p) and $0.05 (720p) per second with audio.

Where to use it: Google Flow, the Gemini API and Vertex AI. One thing to know: video generation inside the Gemini app now runs on Google’s newer Gemini Omni Flash model, not Veo. Omni Flash is built for conversational editing. If you want Veo’s frame control, extension and reference images, go through Flow or the API.

The Anatomy of an Effective Veo 3.1 Prompt

A strong Veo prompt has five layers. You do not need all five every time, but each one you add gives you more control. Google’s own prompt guidance lists the same building blocks: subject, action, style, camera, composition and ambiance.

1. Subject and Action

Who or what is in the frame, and what are they doing? Be specific about appearance, clothing, age and movement.

Weak: “A woman walking” Strong: “A woman in her 30s wearing a long navy wool coat walks through a narrow cobblestone alley, her hand trailing along a weathered stone wall”

2. Environment and Setting

Where does it happen? Include time of day, weather and spatial detail.

Weak: “In a city” Strong: “A rain-soaked Tokyo side street at 2 AM, neon signs reflecting in shallow puddles, steam rising from a street food cart in the background”

3. Camera and Composition

How is the shot framed, and what does the camera do? Veo understands standard film terms.

Weak: (not mentioned at all) Strong: “Slow tracking shot following the subject from behind, gradually pulling wider to reveal the full street, low angle, shallow depth of field”

4. Lighting and Color

Light source, quality and color temperature. This single layer can completely change the mood of the same scene.

Weak: “Good lighting” Strong: “Warm tungsten light spilling from shop windows against cool blue neon reflections, deep shadows between buildings, haze from the rain”

5. Mood and Sound

What should the viewer feel, and what should they hear? With native audio, sound belongs in every Veo prompt.

Weak: “Cinematic” Strong: “Melancholic and quiet. Rain pattering on awnings, a distant train, her footsteps echoing off stone”

Stack all five and Veo has enough to produce a clip that looks intentional instead of random.

How to Prompt the New Veo 3.1 Features

Dialogue and sound effects

Google’s documentation gives two rules: put specific speech in quotes, and describe sound effects explicitly. Say who speaks and how.

Medium close-up of an old fisherman in a yellow raincoat on a rocking boat deck at dawn, grey sea behind him. He squints at the horizon and mutters, "Storm's coming early this year." Waves slapping the hull, gulls crying, the creak of wet rope.

Keep lines short enough to fit the clip. An 8-second clip fits one or two short sentences, not a speech.

Ingredients to Video (reference images)

When you upload up to three reference images, the images carry the look. Your text should describe what the subject does, where, and how it is filmed. Refer to the subject plainly, the same way each time.

The woman from the reference images sits at a small cafe table by a rainy window, stirring her coffee, then looks up and smiles at someone off camera. Medium shot, slow push-in, warm interior light against grey daylight outside. Rain on glass, soft cafe chatter, a spoon clinking on ceramic.

For product shots, upload the product from two or three angles so labels and shapes stay correct, then spend your words on motion and light.

Frames to Video (first and last frame)

With a start image and an end image, your prompt describes the journey between them: the camera move, the action, and how the light changes.

The camera slowly cranes up from the sleeping puppy on the kitchen floor to reveal the whole sunlit kitchen, as morning light sweeps across the tiles. Soft ticking clock, birds outside the window.

Do not describe a scene that contradicts either frame. If the last frame is a wide shot, do not ask for a close-up ending.

Extend

Each extension continues from the final second of your clip. Write the next beat of action, not a new scene. Keep the same character description, setting and sound bed so the join is invisible.

She reaches the end of the alley, pauses under a flickering street lamp, and turns to look back over her shoulder. The camera holds at the same low angle. Rain continues, a dog barks in the distance.

Vertical 9:16

For Shorts, Reels and TikTok, set 9:16 in the settings and compose for a tall frame in the prompt: one subject, centered, with space above and below.

For more examples in the short-form style, see the Veo 3 prompt generator page.

15+ Veo 3.1 Prompt Examples by Category

These work on Veo 3.1 and its Fast and Lite versions. Each follows the layered structure above.

Cinematic and Film

1. Noir Detective Scene

A private detective in a rumpled gray suit sits alone in a dimly lit office, cigarette smoke curling past a half-closed Venetian blind. A desk lamp casts a harsh pool of warm light across scattered papers. Low angle looking slightly up at him. Slow push-in toward his face as he picks up a ringing rotary phone and says, "Yeah. Who's asking?" Black and white, high contrast, deep shadows, film grain. Rain against the window and a ticking clock.

2. Sci-Fi Establishing Shot

A massive generation ship drifts through a nebula of violet and amber gas clouds, its hull covered in geometric light patterns that pulse slowly. Tiny maintenance drones move along the exterior. Extreme wide shot with a very slow dolly from left to right, the ship filling two-thirds of the frame. Cool blue key light from the nebula, warm amber fill from the ship's own lights. Awe-inspiring and lonely. A deep, resonant hum of engines.

3. Period Drama Dialogue

Two women in Edwardian-era white linen dresses sit across from each other at a small wrought-iron garden table, a porcelain tea set between them. One leans forward and says quietly, "You knew, didn't you?" The other looks away toward a distant hedge maze. Medium close-up, shallow depth of field, soft garden bokeh. Late afternoon golden light through a willow tree. Restrained tension. Birdsong and the gentle clink of a teacup on a saucer.

Product and Commercial

4. Luxury Watch Ad

A brushed steel chronograph watch rests on a slab of dark volcanic rock. A thin stream of water flows slowly over the rock, droplets catching light as they pass over the watch crystal. Extreme close-up macro shot with a slow orbit around the watch face, rack focus from the water droplets to the dial. Single overhead spotlight creating sharp highlights on the steel, deep black background. Premium and precise. Quiet ambient tone with soft flowing water.

5. Skincare Product Launch (vertical)

Vertical 9:16 frame. A frosted glass bottle of serum sits on a bed of melting crushed ice, condensation running down the bottle. A single peony petal falls in slow motion past the bottle. Centered product hero shot, slow zoom-in from medium to close-up. Soft diffused overhead light with a cool blue-white tone, clean white background. Fresh and clean. Subtle crystalline chimes.

Tip: upload the real bottle as an Ingredients to Video reference so the label matches your product.

6. Food and Beverage

Hot espresso is poured in slow motion from a brushed copper pitcher into a clear double-walled glass, forming a rich amber crema on top. Wisps of steam rise and catch the backlight. Tight close-up at table level, shallow depth of field. Warm side light from the left glowing through the espresso, dark moody background. Indulgent and inviting. The deep, slowed-down sound of the pour.

7. Doorbell Camera Viral Clip

Doorbell camera footage of a front porch at night, fisheye distortion, grainy infrared black-and-white image, timestamp in the corner. A raccoon climbs onto the porch, looks directly into the camera, then carefully picks up a delivered pizza box and drags it down the steps. Static wide angle, slight motion blur. Include rustling cardboard, tiny claws on wood, and a man's voice from inside yelling, "Hey! That's mine!"

8. Handheld Phone Reaction

Handheld phone video, slightly shaky, in a small apartment kitchen with harsh overhead light. A young man in a gray hoodie opens the fridge and finds his cat sitting inside on the middle shelf, calmly licking its paw. He jumps back and says, "How did you even get in there?" The cat blinks slowly. Vertical framing, natural phone camera quality, fridge hum and a single meow.

9. TikTok Product Showcase

A pair of white wireless earbuds floats against a background that shifts from coral pink to deep purple. The earbuds slowly rotate to show every angle while tiny musical note particles orbit around them. Centered product shot with a slow 360-degree orbit, clean minimal composition. Soft gradient lighting matching the background, subtle rim light on the earbuds. Playful and modern. A light lo-fi beat.

Abstract and Artistic

10. Motion Design Loop

Iridescent glass cubes, spheres and rings slowly assemble into a larger structure, then gently come apart and reassemble in a different shape. Isometric angle, static frame, centered composition. Soft studio lighting with colorful light reflections from the glass onto a matte gray surface below. Satisfying and hypnotic. Gentle chimes that follow the rhythm of the assembly.

11. Nature Timelapse

A single mushroom pushes up through the dark forest floor and unfurls its cap, several hours compressed into 8 seconds. Morning dew collects on its surface while tiny ferns slowly uncurl around it. Extreme close-up macro shot at ground level, locked-off camera. Light rises from deep blue darkness to warm amber, shafts of light moving across the forest floor. Quiet wonder. Birdsong fades in as the light grows.

12. Abstract Data Visualization

Thousands of glowing blue and white particles flow through a dark void, forming the shape of a neural network, synapses firing in cascading waves of brighter light. The structure slowly rotates to reveal its depth. Wide shot pulling back to show the full structure, slow continuous rotation. Self-illuminated, deep black background. Futuristic and awe-inspiring. Deep pulsing synth tones in time with the bursts.

Storytelling and Narrative

13. Opening Title Sequence

A vintage leather journal lies open on a wooden desk. An invisible hand writes elegant cursive with a fountain pen, the words appearing one by one as ink spreads slightly into the paper grain. Overhead close-up looking straight down, static camera with very subtle breathing movement. Warm candlelight from the right, the rest of the room in soft darkness. Mysterious and literary. The scratch of pen on paper, a clock ticking in the distance.

14. Character Introduction

A young chef stands alone in a professional kitchen before the morning rush, running her hand along the stainless steel counter. She ties on a white apron, adjusts her knife roll, closes her eyes and takes one deep breath. Medium shot from across the pass, shallow depth of field, slow handheld tracking as she moves. Cool fluorescent kitchen light mixed with warm sunlight from a high window. Focused, the calm before the storm. Ventilation hum, a radio playing faintly from another room.

15. Emotional Climax

An elderly man sits on a park bench holding a creased photograph, his thumb tracing the face in the image. He looks up slowly as autumn leaves drift across the frame, and a small smile breaks through. Close-up on his face with a slow pull-back to a medium shot that reveals the empty space beside him on the bench. Overcast soft light with warm golden tones from the leaves, slightly desaturated. Bittersweet and tender. Wind through dry leaves, a distant church bell.

16. Two-Shot Dialogue for Extension

Medium two-shot of a teenage boy and his grandmother on a porch swing at sunset. She hands him an old brass key and says, "Your grandfather wanted you to have this." He turns it over in his fingers, surprised. Static camera, warm low sun from the side. Crickets, the creak of the swing.

Then extend with the next beat: “He looks up at her and asks, ‘What does it open?’ She smiles and points toward the old barn behind the house. The camera slowly pans to follow her hand.”

Before and After: Vague vs Structured Veo Prompts

Vague: “A dog running on a beach”

Structured:

A golden retriever sprints along the shoreline of a wide sandy beach, water splashing under its paws with each stride, ears flapping, tongue out. Tracking shot running alongside at the dog's speed, low angle from knee height. Late afternoon sun low over the ocean behind the dog, warm backlight and lens flare. Joyful and free. Waves, paws hitting wet sand, happy panting.

The vague prompt gets a generic dog on a generic beach in flat light from a random angle. The structured one tells Veo the breed, the action details, the camera, the light, the mood and the sound. The model does not have to guess.

Vague: “A city at night”

Structured:

Aerial drone shot descending slowly through fog toward a rain-slicked downtown intersection, traffic lights casting red and green pools on the wet asphalt. Taxis and pedestrians with umbrellas cross the frame. High angle descending toward eye level, anamorphic lens with horizontal flares. Amber sodium streetlights, blue-white headlights cutting through the fog, neon storefront signs. Moody and noir. Distant sirens, tires on wet road, rain on umbrellas.

Common Veo Prompt Mistakes (and How to Fix Them)

1. Being too abstract

“A beautiful cinematic video with amazing visuals” gives Veo nothing to render. “Beautiful” is a judgment, not an instruction.

Fix: describe what beautiful looks like: the light, the palette, the composition, the movement.

2. Cramming too many subjects

“A knight fighting a dragon while a princess watches from a tower and villagers flee a burning market with a wizard casting spells overhead” splits the model’s attention five ways.

Fix: one primary subject and action per clip. Use Extend or separate generations for more beats.

3. Skipping the camera

Describing what is in the scene but not how it is filmed is like a screenplay with no shot directions.

Fix: always include a shot type, a camera movement and an angle.

4. Forgetting sound

Veo generates audio whether you ask or not. If you do not describe it, you get whatever the model guesses.

Fix: end every prompt with the sounds you want, and say “no music” if you want none.

5. Writing too much action for 8 seconds

A single Veo 3.1 generation is at most 8 seconds. Three actions and a speech will not fit.

Fix: give each clip one beginning, middle and end. Use Extend for the next beat.

6. Contradicting your reference images or frames

If your reference image shows a red jacket and your text says a blue coat, the output wavers between them.

Fix: let the images carry appearance, and use the text for action, camera, light and sound.

Writing Better Veo Prompts, Faster

The five-layer structure works, but building a full prompt from scratch for every clip takes time. Most creators need 10 or 20 prompts per session, not one.

That is why I built LzyPrompt. You type your idea in plain words, like “raccoon steals pizza from porch,” and it returns 10 prompt variations. Each one has a camera style, setting, lighting, subject, action beats, small imperfections and sound cues, plus a ready-to-post caption. The prompts are plain text, so you paste them into Veo 3.1 (or any other model) and generate. You can also export all 10 to CSV for bulk work.

It is useful if you are not a cinematographer. Not everyone knows the difference between a dolly and a tracking shot, and you should not need to. Try LzyPrompt free for 7 days.

Quick Reference: Camera, Lighting and Mood Keywords Veo Understands

Camera Shots

Extreme close-up (ECU): eyes, hands, small details
Close-up (CU): a face or single object filling the frame
Medium close-up (MCU): head and shoulders
Medium shot (MS): waist up
Medium wide (MW): knees up
Wide shot (WS): full body with environment
Extreme wide (EWS): landscapes, establishing shots

Camera Movements

Static/locked-off: no movement, tripod feel
Pan: horizontal rotation left or right
Tilt: vertical rotation up or down
Tracking/dolly: camera moves with the subject
Push-in: camera moves toward the subject
Pull-back: camera moves away from the subject
Crane/boom: vertical movement up or down
Orbit: camera circles the subject
Handheld: slight natural shake for a documentary feel
Steadicam: smooth movement while following action

Lighting Styles

Golden hour: warm, low-angle sunlight
Blue hour: cool, soft light before sunrise or after sunset
Rembrandt lighting: a triangle of light on the shadowed cheek
High key: bright, even, few shadows
Low key: dark, dramatic, heavy shadows
Backlit/silhouette: light behind the subject
Practical lighting: light from lamps or screens visible in the scene
Volumetric: visible beams through fog, dust or haze
Neon: colorful light from signs
Candlelight: warm, flickering, intimate

Mood Keywords

Epic, grand: dramatic scale
Intimate, tender, quiet: emotional closeness
Tense, suspenseful, uneasy: thrillers
Dreamlike, ethereal, surreal: abstract work
Gritty, raw, documentary: realism
Nostalgic, bittersweet: emotional depth
Energetic, explosive: action
Serene, calm, meditative: peaceful scenes
Eerie, haunting: horror or mystery

For more camera vocabulary, see the camera movement cheatsheet and the lighting prompts guide.

Frequently Asked Questions About Veo 3 and Veo 3.1 Prompts

Do Veo 3 prompts work on Veo 3.1?

Yes. The structure is the same: subject, action, setting, camera, light, mood and sound. Veo 3.1 follows prompts more closely and adds reference images, first and last frames, extension and 9:16 output, so you have more ways to control the result.

How long should a Veo 3.1 prompt be?

Around 50 to 150 words works well for an 8-second clip. Under 20 words and you leave most choices to the model. Much longer and instructions start to compete with each other.

How do I make characters talk in Veo 3.1?

Put the exact line in quotes and say who says it and how, for example: The old man mutters, “Storm’s coming.” Keep it short enough to fit the clip length, and describe background sound separately.

Does Veo 3.1 support negative prompts?

Google’s Veo prompt guidance focuses on describing what you want, not listing exclusions, so treat negatives as a last resort. Describe what should be in the frame instead. If an unwanted element keeps appearing, be more specific about what should fill that space, and say “no music” or “no text on screen” plainly when needed.

How long can a Veo 3.1 video be?

A single generation is 4, 6 or 8 seconds. With Extend in the Gemini API you can add 7 seconds at a time, up to 20 times, for a maximum of 148 seconds at 720p.

Can I use the same prompt in Veo and other models?

Mostly, yes. A well-structured prompt works in Kling, Seedance, Hailuo and others too, though each model has strengths. Kling handles multi-shot sequences in one generation, and Seedance is built around reference files. See our Kling 3.0 prompt guide, the Seedance prompt guide and the 2026 comparison of AI video generators. If you are moving from Sora, which OpenAI has shut down, read the Sora shutdown alternatives guide.

Start Creating Better AI Video

The gap between forgettable and impressive AI video is almost always in the prompt. Use the five layers, add sound to every prompt, and use the Veo 3.1 tools (reference images, frames and Extend) when you need consistency. Start with any of the examples above, change the details for your project, and build your own library from what works.

For more guides, browse the LzyPrompt blog.