Updated November 2026

Veo 3.1 prompting guide: structure and real examples

Write a Veo 3.1 prompt like a shot description: camera and framing, subject, action, setting, style and light, then the sound — dialogue in quotes, effects and ambience on their own line. Set aspect ratio, length and resolution in the settings; text in the prompt does not change them.

Key takeaways

  • Lead with the shot: "Medium shot", "POV shot", "Static camera", "slow push-in".
  • One subject, one main action per clip. A 4–8 second shot has room for one idea.
  • Sound is part of the prompt: Audio: lines for effects and ambience, exact dialogue in quotation marks.
  • Settings beat text: our gallery shows prompts asking for "vertical 9:16" or "10–12 seconds" still came out at the chosen aspect ratio and length.

Try a prompt from this guide

Paste it into the generator, choose Lite, Fast or Quality, and see the credit cost before you generate.

Sign in with Google to create an account; buy credits only when you need them.

The structure: five parts, then sound

Google's official Veo 3.1 prompting guide gives a five-part formula: [Cinematography] + [Subject] + [Action] + [Context] + [Style & Ambiance]. It treats sound as a separate layer with three parts: dialogue in quotation marks, sound effects, and ambient noise. In practice, the strongest prompts in our gallery follow the same order:

PartWhat to writeFrom a real gallery prompt
Camera and framingShot size, angle, movement, lens“Static camera”, “POV shot”, “Camera pushes in slowly and smoothly”
SubjectWho or what, with two or three concrete details“a cartographer in a cluttered study”
ActionOne main thing that happens, in order“slides into a damp running shoe … pulls the foot back out”
SettingPlace, time of day, background“a dim hallway … Rainy day, grey natural light from a window”
Style and lightLook, colour, lighting, film or footage type“Realistic amateur footage, shot on iPhone, no color grading”
SoundDialogue in quotes; effects; ambience“Audio: rain pattering outside, a frustrated sigh”

For things you want left out, Google's guide says to describe the exclusion concretely: specify "a desolate landscape with no buildings or roads" instead of "no man-made structures". Our generator has an Enhance button that rewrites your prompt in clearer, more specific wording while keeping your scene and dialogue (you can undo it), and the prompt library has more starting points.

Settings are not part of the prompt

We measured the actual video files behind several popular gallery prompts. Wherever the text and the settings disagreed, the settings won:

  • The glass bottle prompt ends with "Vertical 9:16" and "4K"; the file is 1920×1080, landscape.
  • The product-problem prompt opens with "4-second vertical video" and a 0–4 s timeline; the file is 8 seconds long (vertical, because 9:16 was selected).
  • The anatomical model prompt asks for a "10-12 second loopable clip"; the file is 8 seconds, the longest single clip Veo 3.1 generates.

So pick 16:9 or 9:16, 4, 6 or 8 seconds, and the resolution in the generator's Aspect Ratio, Duration and Resolution controls (or options.aspectRatio, options.duration and options.resolution in the API), and use the prompt words for what is in the frame. If you write a timeline, make it fit the length you selected.

Real prompts from our gallery

These are prompts imaginary.app users shared publicly, quoted as written (long ones trimmed with …). Each links to the video so you can judge the result yourself. We checked every file for a sound track, and the credits are what the same settings cost today, with audio on or off to match the file.

Two sentences for a 4-second shot

Two warriors crouched above a river, watching a distant tribe across the water. Audio: slow river flow, wind through grass, faint distant voices, low ominous hum.

Veo 3.1 Quality · 4 s · 720p · 9:16 · no sound track · 7 credits · Watch the video

Two sentences can be enough for a 4-second shot: one for the picture, one for the sound. But this shared file has no sound track: either With Audio was off or the model returned a silent clip. An Audio: line describes the sound you want; it does not switch sound on.

One line of dialogue, in quotes

A medium shot, historical adventure setting: Warm lamplight illuminates a cartographer in a cluttered study, poring over an ancient, sprawling map spread across a large table. Cartographer: "According to this old sea chart, the lost island isn't myth! We must prepare an expedition immediately!"

Veo 3.1 Fast · 8 s · 1080p · 16:9 · with sound · 10 credits · Watch the video

Shot type first, then light, subject and action, then the exact words in quotation marks. Google's guide attributes dialogue in prose (A woman says, "…"); this prompt uses a speaker label ("Cartographer:") instead, and the file does have speech. The line is short enough to be spoken inside 8 seconds. The same prompt was also run on Veo 3.1 Lite at 720p, which makes a useful side-by-side if you are choosing a model.

Phone-footage realism for ads

smartphone video, UGC style, handheld with slight natural shake. Close-up of soaking wet hiking boots being dropped onto a doormat in a dim hallway, water dripping off the soles, muddy laces. Rainy day, grey natural light from a window. The camera tilts up slightly as a hand reaches down and squeezes the wet boot tongue, water seeping out. Realistic amateur footage, shot on iPhone, no color grading, slightly overexposed window. Audio: rain pattering outside, a frustrated sigh, the wet squelch of the boot.

Veo 3.1 Lite · 4 s · 720p · 9:16 · no sound track · 1 credits · Watch the video

It names the camera style before anything else ("handheld with slight natural shake") and gives one camera move. Describing imperfections — overexposed window, no colour grading — is how you ask for a raw, unpolished look. Its Audio: line lists three sounds, but this shared file is silent.

A POV shot with one action and a reaction

Vertical 9:16 smartphone video, UGC style. POV shot: a foot in a white sock slides into a damp running shoe by the front door, early morning light. The person pulls the foot back out immediately in discomfort. Handheld, realistic amateur footage, shot on iPhone, natural colors, shallow depth of field. Audio: the damp fabric sound, a disgusted "ugh" grunt, birds chirping faintly outside.

Veo 3.1 Quality · 6 s · 1080p · 9:16 · no sound track · 10 credits · Watch the video

A clear point of view and a single action with a reaction. The prompt writes a non-verbal sound in quotes ("ugh"), but the shared file has no sound track. Note that "Vertical 9:16" in the text is not what made this clip vertical; the aspect-ratio setting was (see the section on settings below).

Abstract visuals with a sound design brief

Cinematic vertical 9:16 video, 8 seconds. Black smoke drifts slowly. Gold embers float over a pure black void. A single soft god-ray cuts through. Tiny gold particles rise, resembling money. Shallow depth of field. Dust drifts. Colors are deep black, charcoal, with a hint of gold. Camera pushes in slowly and smoothly. Audio: A tense cinematic riser leads into a single deep sub-bass boom on the first frame. A low pulsing electronic heartbeat drone begins. Soft shimmering chimes and 'cha-ching' cash-register accents punctuate the sound. Ambient room tone. No spoken dialogue. Negative: No text, numbers, captions, letters, UI, watermarks, or logos.

Veo 3.1 Lite · 8 s · 1080p · 9:16 · with sound · 5 credits · Watch the video

Short declarative sentences, a fixed colour palette, one camera move, and a timed audio cue ("on the first frame"). "No spoken dialogue" and the closing list of things to leave out are aimed at stray voices and garbled on-screen text, a common Veo failure covered in our gibberish subtitles fix.

Motion graphics with exact counts

A clean, elegant motion graphics timeline animation on a soft dark background (deep navy or charcoal, not pure black). A thin, smooth horizontal line stretches across the frame in a warm gold or soft teal color. Exactly four small circular dots are evenly placed along the line — no more, no fewer. A soft glowing dot travels gently and smoothly along the line, moving from the first point to the second, pausing briefly, then continuing to the third, and finally the fourth. … Muted, sophisticated color palette, gentle easing motion, no harsh flashes or sparks, flat modern vector design. Calm and refined pacing throughout.

Veo 3.1 Fast · 6 s · 1080p · 16:9 · no sound track · 5 credits · Watch the video

For graphics, state counts and order explicitly ("exactly four … no more, no fewer"), describe the motion as a sequence, and name the easing and pacing. Another gallery prompt describes a high-energy version of a similar line-and-nodes graphic; compare the two to see what the mood words change.

A structured (JSON) prompt with timed beats

{ "prompt": "a minimalist 2d whiteboard animation. pure white background. pure pure black thin hand-drawn lines. bright yellow accents only for key details. at the bottom of the screen, big bold black marker text displays subtitles.", "sequence": [ { "time": "0-3 seconds", "action": "a minimalist stickman drawn with black lines stands next to a large downward arrow. …", "subtitle_text": "market is down again" }, { "time": "3-6 seconds", "action": "hard cut. stickman sits at a desk, typing on a laptop. …", "subtitle_text": "but don't panic" }, { "time": "6-9 seconds", "action": "hard cut. close up. stickman holds a heavy bag with a yellow dollar sign. …", "subtitle_text": "it's time to buy" } ] }

Veo 3.1 Fast · 8 s · 1080p · 16:9 · with sound · 10 credits · Watch the video

Veo reads JSON as text, so this works as a tidy way to separate style from timed beats. Two cautions: the sequence runs to 9 seconds but the clip is 8, so the last beat gets squeezed; and on-screen text is the part most likely to come out misspelled, so check the frames before you publish.

Physics, step by step, with a locked camera

A giant transparent glass bottle filled with bright red soda is resting on the first stair. No people. No hands. No human body parts visible. The bottle is released and begins rolling DOWN THE ENTIRE STAIRCASE. The camera remains fixed at the top, watching the bottle travel downward. The bottle hits the first stair and bounces. … After traveling approximately halfway down the staircase, the bottle strikes a stair edge with extreme force. Only then does the glass bottle shatter into hundreds of realistic glass fragments. … Photorealistic physics, realistic gravity, realistic glass destruction, realistic fluid simulation, cinematic slow motion at the moment of breaking, viral Instagram reel style, outdoor daylight, ultra realistic 4K. Vertical 9:16.

Veo 3.1 Fast · 8 s · 1080p · 16:9 · no sound track · 6 credits · Watch the video

Breaking an event into ordered beats and fixing the camera ("remains fixed at the top") gives the model a clear plan. But this prompt also shows the limits of text: it ends with "Vertical 9:16" and "ultra realistic 4K", yet the file is a 1920×1080 landscape video, because the settings said 16:9 and 1080p.

For more in this style, see cinematic and product shot prompts and dialogue and sound prompts, or browse the full gallery.

Writing dialogue and sound

  • Put the exact words in quotation marks and say who speaks in prose, as Google's guide does: A woman says, "We have to leave now." Some gallery prompts use speaker labels ("Cartographer:") instead; Google's guide doesn't mention them.
  • Keep a line short enough to say inside the clip. One or two sentences fit in 8 seconds.
  • Describe effects and ambience separately ("SFX: thunder cracks in the distance", "Ambient noise: the quiet hum of a starship bridge", both from Google's guide).
  • If you don't want speech, say "No spoken dialogue"; if you don't need sound at all, switch audio off, which costs fewer credits.

If a clip comes back silent or with the wrong sound, our Veo audio troubleshooting guide covers the usual causes.

What gets prompts rejected

There are two layers of checks, and it helps to know which one stopped you:

  • Our pre-check runs before any credits are taken. It refuses prompts asking for deepfakes, face swaps, impersonating or cloning a real person's face or voice, or non-consensual intimate imagery. The API returns these as CONTENT_POLICY_VIOLATION.
  • Google's safety filters check the prompt and the finished video. Google lists categories including celebrity (a photorealistic depiction of a prominent person), child, sexual, violence, hate, dangerous content and third-party content. We never enable Google's setting that allows depicting children, so those prompts are refused. If Google blocks a generation, the credits are returned automatically.

Usually the fix is to describe a fictional person instead of naming a real one, and to describe the scene you want instead of piling up forbidden words. Our "something went wrong" error guide goes through the error messages in detail.

Choosing aspect ratio, length and model

Use 9:16 for Shorts, Reels and TikTok (see social media shorts) and 16:9 for YouTube, websites and presentations. Use 4 seconds for a single action or a reaction, 6 for a short move, and 8 for dialogue or a sequence of beats. For anything longer, plan several clips; this guide to the 8-second limit shows how.

A cheap way to iterate (credits, with audio)
StepSettingsCredits
Draft the ideaVeo 3.1 Lite, 720p, 4 s2
Final, most usesVeo 3.1 Fast, 1080p, 8 s10
Final, highest qualityVeo 3.1 Quality, 1080p, 8 s26

Frequently asked questions

What is the best prompt structure for Veo 3.1?

Google's own guide suggests cinematography, subject, action, context, then style and ambiance. Add sound last: dialogue in quotation marks, sound effects and ambient noise on their own lines. Put aspect ratio, length and resolution in the settings, not the prompt.

How long can a Veo 3.1 prompt be?

On imaginary.app, up to 2,000 characters in both the generator and the API. Most of the gallery prompts above are well under that; longer is not automatically better.

Can I write "9:16" or "10 seconds" in the prompt?

You can, but it does not change the output. In our gallery, prompts that asked for vertical video or 10–12 second clips produced whatever the aspect-ratio and duration settings said. Choose 16:9 or 9:16 and 4, 6 or 8 seconds in the settings.

Why was my prompt rejected?

imaginary.app blocks prompts asking for deepfakes, face swaps, impersonation of real people or non-consensual imagery before any credits are used. Google's own filters can also block prompts or finished videos, for example photorealistic depictions of prominent people or of children. If Google blocks a generation, your credits are returned automatically.

Does the same prompt work in Google Flow?

Yes. Flow, the Google Cloud API and our generator run the same Veo 3.1 models (in Flow, only Lite after December 5, 2026), so the same prompt structure applies. Results vary between runs and between Lite, Fast and Quality.

How we checked: prompts, models and lengths come from our public gallery listing on the date below; we read each video file's resolution, duration and sound track with ffprobe. Prompt advice attributed to Google is quoted from Google's guide.

Sources

Prices, model specs and dates on this page were checked on November 8, 2026. Third-party prices change; check the source before you rely on a number.

Write your first Veo 3.1 prompt

Text-to-video, image-to-video and first-and-last-frame on Lite, Fast and Quality. Credits come back automatically if a generation fails.

Sign in with Google to create an account; buy credits only when you need them.