Prompt guide

Veo 3.1 Prompts: How to Write for Picture and Sound

Published October 7, 2026By DreamClip

Good Veo 3.1 prompts are short, ordered and written for two senses. Veo 3.1 renders the sound with the picture, so a prompt that names the shot, the action, the light and one or two sounds comes back as a clip with a soundtrack that fits it. A prompt that lists ten adjectives and no sound comes back as a mood board with a hum.

This guide is the method: the order that works, the audio line, camera words, ads, image to video, twelve prompts that made real Veo clips (eleven of them on our Veo 3.1 page), and the failures you will hit.

On DreamClip, both Veo 3.1 (with audio) and its quicker sibling, Veo 3.1 Fast, render from a text prompt or a start frame at 720p or 1080p, 16:9 or 9:16. Both open in the Video studio with the model already picked, and a Veo run that fails is not billed.

Veo 3.1 prompts: the anatomy of one that works

A video prompt is a shot list for one shot, and the early words carry the most weight, so the order below is the control you have.

Shot and camera: close-up, handheld, slow push in.
Subject and action: a baker scores a loaf with one quick cut.
Setting and light: a small bakery before dawn, one warm lamp over the bench.
Sound: the blade through the crust, an oven ticking as it heats.
Look: photoreal, shallow depth of field, 9:16.

Put together, that is about 40 words, plenty for an 8 second clip. The two parts people leave out are the action and the sound: without a verb the shot drifts, and without a sound line the soundtrack turns generic.

The two Veo models and what the studio offers

Two Veo models run on DreamClip. The cells are read off our catalog: what the studio offers today, not Google's full specification.

Read from the DreamClip catalog when this page was built.
Model	Maker	Clip lengths	Sizes	Sound	Start frame	Reference photos
Veo 3.1 (with audio)	MakerGoogle	Clip lengths4, 6 or 8 s	Sizes720p, 1080p	SoundYes, on by default	Start frameYes	Reference photosNo
Veo 3.1 Fast	MakerGoogle	Clip lengths4, 6 or 8 s	Sizes720p, 1080p	SoundYes, on by default	Start frameYes	Reference photosNo

Veo 3.1 (with audio) is the full model; Veo 3.1 Fast is the same family at a lower cost per second and is where drafts belong. The two cards below play clips each Veo model rendered itself.

Neither Veo model takes an end frame or reference photos here, and neither renders square or 4K. A clip that ends on a second image belongs on Seedance 2.5, which has start and end frames; native 4K belongs on Kling 3.0 4K. The Veo 3.1 vs Kling 3.0 comparison says when each switch pays off.

Writing the audio line: dialogue, ambience and music cues

Veo 3.1 audio is generated in the same pass as the picture, on by default in the studio. The prompt is the only place you direct it, so give it one specific line: the source, the space, and what stays quiet.

One line of sound per prompt. More than three sources and the mix turns to mush.
Sound you want	How to write it	Example
A spoken line	How to write itWho speaks, then the words in quotes	ExampleA man at the ticket window says "Two for the late show."
Room tone	How to write itName the space; the echo follows	ExampleA tiled kitchen; an empty car park
Effects	How to write itTwo or three sources, in order	ExampleRain on a window, a kettle starting to whistle
A music cue	How to write itInstrument, mood, level	ExampleA single piano, slow, low under the voice
Silence	How to write itSay what not to add	ExampleNo music. No voices. Only the sea.

For dialogue, keep the line under ten words for a 6 or 8 second clip and give the speaker something to do while saying it. For a voiceover you will record yourself, write "no music, no voices, room tone only".

Camera language Veo 3.1 understands

Veo follows camera instructions when they are film words, not app words. Say the move and, if it matters, the lens. One move per clip.

Say this	What you get	Good for
Slow push in	What you getThe camera creeps toward the subject	Good forReveals, faces, product details
Slow orbit	What you getThe camera circles the subject	Good forProducts, a frozen moment, a dancer
Tracking shot	What you getThe camera moves with a moving subject	Good forRunning, cycling, a dog on a beach
Handheld	What you getSmall natural shake	Good forVlogs, kitchens, anything candid
Low wide angle	What you getLooking up, exaggerated scale	Good forGiants, buildings, drama
Aerial, descending	What you getA drone drop toward the ground	Good forLandscapes, cities, water
Macro	What you getExtreme close-up, shallow focus	Good forWatches, liquid, texture
Locked off	What you getNo camera movement	Good forComedy, a single beat, a logo

Two words to handle with care: "zoom" on its own, which is ambiguous between a lens and a crop, and "cinematic" on its own, which says little. Spell out the path instead, as the Earth prompt below does with "accelerating up through clouds as the city shrinks". "35mm, shallow depth of field" tells it more than any adjective.

Prompts for ads and product shots

Veo 3.1 for ads is strongest when the sound is part of the sell: a pour, a click, a line that lands the tagline. Write the product as the subject, with one action, one light and one sound.

The reveal: "Macro, slow push in: a matte black earbud case opens on dark stone, a soft click, a thin ring of light inside, no music."
The spokesperson: "Medium shot, handheld: a courier in a yellow jacket taps a parcel and says “Same day, every day,” a city street behind, traffic and a passing bus, 9:16."
The pour: "Close-up: cold brew pours over clear ice in a tall glass on a sunlit counter, condensation forming, the pour and the ice cracking, photoreal."

For a product that turns a full circle, or a clip past 8 seconds, Kling 3.0 is the better tool; the Kling page collects product prompts for it. Our text to video page has more ad-ready prompts.

Image to video prompts: describe the motion, not the picture

With a start frame attached, the subject, the setting and the light are already in front of Veo 3.1, so a prompt that describes them again wastes its attention. Write only what changes: the movement, the camera, the sound. Open Veo 3.1 with a start frame, upload the photo, and keep the prompt to a sentence or two.

A woman on a balcony: "She turns to look over her shoulder and smiles, wind in her hair, a slow push in, street noise far below."
A product on a plinth: "A slow orbit, the light sweeping across the surface, a low hum, nothing else moves."
An old family photograph: "The people breathe and shift slightly, a curtain moves behind them, soft room tone, the camera holds still."

Match the shape to the photo, 9:16 for a portrait and 16:9 for a landscape, and keep the first second calm. The image to video page covers the whole method.

Twelve Veo 3.1 prompts that made real clips

The first eleven are Veo 3.1 renders shown on our Veo 3.1 page with their prompts; the last made the Veo 3.1 Fast preview above. Read them for the pattern: one subject, one action, a camera, a light.

Quoted as written from our Veo 3.1 page and the Veo 3.1 Fast preview.
Prompt	What to notice
A border collie confidently rides a turquoise ocean wave on a surfboard, spray backlit by the sun, tongue out, slow motion water droplets, summer adventure energy, vertical composition	Fur, water and anatomy in one impossible shot
A joyful giant man in casual clothes strides down a downtown avenue standing waist-high to the skyscrapers, crowds cheering from sidewalks below, confetti drifting between buildings, dramatic low wide angle, playful epic scale, vertical composition	Scale set by a comparison and a camera angle
An elegant woman in a shimmering gold dress stands beside a three-tier birthday cake as cold-spark fountains erupt around it, a champagne tower glittering behind her, gold confetti drifting down, ballroom bokeh lights, slow cinematic push-in, luxurious celebration, vertical composition	Several lights named, one camera move
Bullet-time camera orbit around a dancer frozen mid-leap inside a burst of suspended water droplets, studio strobes glinting through the drops, time snaps back to motion at the end, dramatic dark stage, vertical composition	A time instruction with a beginning and an end
Molten liquid chrome rises from a black mirror floor and sculpts itself into a floating abstract triangular emblem, ripples settling, crisp studio reflections, dark premium backdrop, slow reveal, vertical composition	An abstract subject with physical verbs
Intimate close-up portrait in warm candlelight: a young woman closes her eyes, makes a wish and softly blows out the candles on a birthday cake, tiny embers drifting, fairy-light bokeh behind her, gentle slow motion, heartfelt and warm, vertical composition	Three small actions in order; the quiet shots look most real
Night streetball: a basketball swishes through a chain net on an outdoor city court under floodlights as a simple digital shot clock hits zero, a crowd of friends along the chain-link fence erupts in celebration, dramatic backlit haze, gritty triumphant slow motion, vertical composition	A sound source written as an action
A fluffy orange cat wearing a small chef hat stands at a tiny stove and flips a mini pancake in a pan, a puff of flour in the air, cozy warm kitchen light, shallow depth of field, charming and absurd, vertical composition	Comedy played straight, one physical beat
Hyper zoom-out beginning on a girl waving from a rooftop, accelerating up through clouds as the city shrinks to coastline, continents, then the glowing curve of Earth in starfield space, seamless continuous pullback, vertical composition	One continuous move described as a sequence
A glowing neon tube light traces itself into an abstract geometric diamond emblem on a dark brick wall, electric glow blooming, faint hum flicker, camera slowly pushes in as the sign snaps fully lit, moody night studio, vertical composition	The sound and the ending both written
An astronaut holds a selfie stick and vlogs while floating in a space station cupola, Earth turning slowly below, waves at the camera with a grin as sunrise flares over the horizon, authentic handheld vlog feel, vertical composition	Handheld, with a light change inside the clip
A border collie leaps for a frisbee on a windy beach, sand spraying, ears flying, the wave line breaking behind, gulls and surf audible, slow motion catch, tracking shot, photoreal	The Veo 3.1 Fast preview: sound and camera both named

Common failures and the fix

What went wrong	Why	The fix
Nothing moves	WhyNo verb in the prompt	The fixOne action for the subject, one move for the camera
The wrong thing moves	WhyTwo subjects, two actions	The fixCut to one subject; the second becomes scenery
Lips do not match the words	WhyA long line, or nothing for the speaker to do	The fixUnder ten words, in quotes, with an action during the line
Generic music over everything	WhyNo sound line, so Veo scores it	The fixWrite the sound you want, or "no music"
The clip ends too early	WhyAn 8 second idea in a 4 second clip	The fixPick 8 seconds, or cut the idea to one beat
Hands and small objects warp	WhyFast action close to the lens	The fixSlow the action or pull the camera back
Text in the frame is gibberish	WhyWords in a moving shot rarely hold	The fixKeep words out of the shot; add titles in an editor

When a prompt keeps missing, change one thing at a time and re-run it on Veo 3.1 Fast, which costs less per second.

Prompt length and the 4, 6 or 8 second choice

The studio box takes up to 4,000 characters for Veo 3.1 as of October 2026, and you should never need a tenth of that. Forty to seventy words is where the twelve prompts above sit. In a long prompt the last instructions are the likeliest to be dropped, and that is usually the sound line.

Both Veo models render 4, 6 or 8 seconds, priced by the second, so draft short.
Length	Fits	Example beat
4 s	FitsOne reaction, one reveal, one logo settle	Example beatThe earbud case opens and clicks
6 s	FitsOne action with a lead-in	Example beatShe turns, smiles, the camera pushes in
8 s	FitsSetup, move and payoff, or one spoken line	Example beatThe courier taps the parcel and says the line

Pick the length from the idea. A 4 second idea stretched to 8 drifts; an 8 second idea cut to 4 ends mid-move. Past 8 seconds, cut two clips together or use a model that runs longer.

How to run a Veo 3.1 prompt on DreamClip

Open the Video studio on Veo 3.1 Fast for drafts, or Veo 3.1 for the final. Each Veo card shows its own starting price.
Pick the shape, the size and the length: 16:9 or 9:16, 720p or 1080p, 4, 6 or 8 seconds. Leave sound on unless you will score the clip yourself.
Paste the prompt in the order above. To attach a photo, switch to Frame to Video, add it as the start frame, and describe only the motion.
Run it. Credits are held at submit and settled only when the clip arrives; a failed run releases them.

To test a prompt against another family, open Compare and pick Veo 3.1 (with audio) beside up to three other models. Compare keeps every lane short and small, in 16:9, and nothing is charged for a lane that does not deliver. The video templates page has more prompts ready to run.

Where Veo 3.1 falls short

Veo 3.1 on DreamClip stops at 8 seconds, renders 720p or 1080p rather than the 4K Google lists on its own page (checked October 2026), takes no end frame and no reference photos, and renders only 16:9 and 9:16. It also costs more per second than Veo 3.1 Fast. On the prompt side it rarely holds legible text, it struggles with fast hands close to the lens, and it simplifies a crowded action to keep the mood you asked for.

None of that is a reason to avoid it. Write the five lines, pick the length from the idea, draft on Veo 3.1 Fast and render the keeper on Veo 3.1.

Questions

How long can a Veo 3.1 prompt be?

The studio box takes up to 4,000 characters as of October 2026. Shorter does better: forty to seventy words with one subject, one action, a camera move, a light and a sound line.

Does Veo 3.1 generate audio from the prompt?

Yes. Sound comes with the picture on Veo 3.1 and Veo 3.1 Fast, on by default in the DreamClip studio. Write the sound you want in the prompt, or switch it off.

Can Veo 3.1 do dialogue?

Yes. Say who speaks and put the words in quotes, keep the line under ten words for a 6 or 8 second clip, and give the speaker an action during the line.

Can I use Veo 3.1 for image to video?

Yes. Open Veo 3.1 with a start frame in the Video studio, upload the photo and describe only the motion. Veo 3.1 takes a start frame but not an end frame on DreamClip; Seedance 2.5 takes both.

What is the difference between Veo 3.1 and Veo 3.1 Fast?

Veo 3.1 Fast has the same lengths, sizes, shapes and sound as Veo 3.1 at a lower cost per second. Draft there and render the keeper on Veo 3.1.

Can Veo 3.1 put readable text in a clip?

Rarely. Words on a sign or a label tend to warp once the camera moves. Keep text out of the prompt, render the shot clean and add titles in an editor afterwards.

Sources

Google DeepMind, the Veo model page, checked October 7, 2026