Clipwave
Media ForgeGenerate images & video, fastMedia StudioNewBuild images & video by chatting with AIWorkflowsAutomate pipelines at scaleClipperNewBreak a video down & cut the highlightsMarketing AgentDone-for-you UGC video ads
Explore NewMCPPricingFAQ
Log inStart free
AI Models/Veo 3.1/Prompting

How to prompt Veo 3

Veo 3.1 reads a prompt the way a director reads a treatment: a 100-150 word paragraph that declares the style first, then camera, subject, setting, light, action and sound. It generates synchronized audio natively — dialogue, ambience, sound effects and music are all asked for inside the prompt. Short keyword prompts produce flat, generic footage; a structured prose paragraph with one clear camera instruction is what unlocks the cinematic look this model is known for.

Anatomy of a Veo 3.1 prompt

  1. 1

    Style declaration

    Open with the look before any detail: "Cinematic product commercial, shot on 35mm", "UGC-style handheld phone footage", "Nature documentary". If you skip it, the model picks a style at random — and it will differ between takes.

  2. 2

    Camera & lens

    One shot type, one movement: "slow dolly-in from a wide to a medium", "static tripod shot", "handheld tracking". Add a focal length ("85mm, shallow depth of field") when you want compression or bokeh.

  3. 3

    Subject

    A detailed physical description — age, clothing, texture, material. Veo renders what you specify and improvises what you leave out.

  4. 4

    Setting & light

    A specific, sensory place plus the light source and direction: "a mountain cabin at dawn, mist rolling through pines, low golden light raking across the deck". Light words do more for the cinematic feel than any quality adjective.

  5. 5

    Action beats

    What happens, beat by beat, in order: "She lifts the mug. Steam curls up. She looks out the window." Complex scenes need the action written frame-by-frame or the model compresses it.

  6. 6

    Audio cues

    Describe the soundtrack explicitly: "Sound of gentle water drops", "Upbeat electronic music", ambience, and dialogue with a named speaker ("the barista says: …"). No audio cues means generic sound — or near-silence.

Template

[style declaration], [shot type + lens + camera movement]. [detailed subject] in [specific setting], [lighting quality and direction]. [Action beat 1]. [Action beat 2]. Audio: [dialogue in quotes / ambience / music cue].

11 example prompts that work

Cinematic establishing shot

16:9 · 8s · 1080p · audio on
Cinematic establishing shot on 35mm film. Slow dolly-in toward a timber mountain cabin at dawn, mist rolling through the pine forest behind it, warm light glowing in one window, frost on the deck railing catching the first sun. The camera pushes in steadily as a thin line of chimney smoke drifts up. Audio: soft wind through pines, distant birdsong, a low ambient drone that swells slightly as the cabin fills the frame.

Style first, one camera move, one micro-event (the chimney smoke) to give the shot life. The swelling drone cue times the music to the dolly — Veo syncs audio to described motion remarkably well.

UGC-style ad with a spoken line

9:16 · 8s · 720p · audio on
UGC-style handheld phone footage, selfie framing with slight sway. A woman in her late 20s with curly hair and a grey hoodie sits in her car, daylight through the windshield, and holds a small amber supplement jar up next to her face. She says with genuine surprise: "I did not expect this to actually work — week two and I can feel the difference." She taps the lid twice on the last word. Audio: quiet car interior, faint traffic outside.

Declaring "UGC-style handheld" up front sets grain, framing and imperfection in one move. Attributed quoted dialogue is lip-synced natively; the lid tap gives the line a visual button so the clip does not end on dead air.

Product commercial hero shot

16:9 · 6s · 1080p · audio on
Cinematic product commercial. Static tripod shot, macro lens. A matte white smartwatch rotates slowly on a brushed-steel turntable against a graphite gradient background, a single softbox key light from the upper left carving the case edge, a thin rim light tracing the crown. Halfway through, the screen wakes and casts a faint glow on the strap. Audio: minimal electronic music with a soft synth hit exactly when the screen lights up.

The screen-wake beat gives the sound design an anchor — Veo lands the synth hit on the visual event. "Static tripod shot" is how you lock the camera; there is no camera-lock setting.

Dialogue scene

16:9 · 8s · 1080p · audio on
Cinematic drama, warm practical lighting. Medium two-shot at eye level in a small bookshop after closing, dust motes in lamplight. An older man in a cardigan hands a wrapped book across the counter to a young woman in a rain jacket. He says quietly: "She wanted you to have this one." The woman looks down, then back up, and answers: "Then I will read it twice." Audio: rain against the shop window, the paper wrapping crinkling, no music.

Name each speaker by description and give each a short line — Veo attributes the lip-sync to the right face. "No music" is a valid audio cue: silence under rain does the emotional work here.

Image-to-video: animate a generated still

i2v · 9:16 · 6s · 720p
Animate this frame: the model holds her pose while a light breeze moves her hair and the silk scarf, the neon sign behind her flickers twice, shallow depth of field holds as the camera drifts a few centimeters to the right. Audio: distant city hum, a faint neon buzz.

In image-to-video your uploaded frame carries composition, wardrobe and light — the prompt directs only motion and sound. Generate the still with an image model in Media Forge first and you control the exact opening frame of the ad.

Nature documentary moment

16:9 · 8s · 1080p · audio on
Nature documentary style, telephoto lens with compressed background. A red fox steps carefully across fresh snow at the edge of a birch forest, ears rotating as it listens, breath visible in the cold air. It pauses, tilts its head, then pounces nose-first into the snow. Overcast soft light, muted whites and greys. Audio: sound of gentle wind, snow crunching under paws, a muffled whump on the pounce — no music, documentary ambience only.

Veo's physics carries this shot: visible breath, snow compression, the pounce arc. Foley described per action ("crunch", "whump") lands in sync — write sound effects as events, not as a mood.

Vertical hook for Reels

9:16 · 8s · 720p · audio on
UGC-style vertical video, punchy and fast. Extreme close-up of a chef's knife halving a dragon fruit in one clean stroke on a wooden board, juice beading on the blade — then a whip-pan up to the grinning chef who says: "You have been cutting these wrong your whole life." Hard overhead kitchen light. Audio: the crisp slice, a whip-pan whoosh, then his line over faint kitchen clatter.

The first two seconds are the hook — lead with the most arresting action, not context. The whip-pan is both the transition and the audio cue; Veo stages all of it inside one generation.

Night city mood piece

16:9 · 8s · 1080p · audio on
Moody cinematic footage, anamorphic lens. Slow lateral tracking shot along a rain-soaked Hong Kong street at night, neon signs in pink and cyan reflecting in puddles, a lone street vendor steaming dumplings under a tarp, passersby with umbrellas blurring past in the foreground. Palette anchored in teal, magenta and sodium orange. Audio: rain on tarpaulin, sizzling steam, muffled Cantonese chatter, a slow synthwave pad underneath.

Naming three palette anchors steers the grade harder than "beautiful neon night". Foreground blur from passersby is a composition instruction Veo executes — it sells the tracking shot depth.

Action with real physics

16:9 · 6s · 1080p · audio on
Cinematic sports footage, low-angle handheld tracking. A runner in a soaked red windbreaker sprints through an empty parking garage during a storm, footfalls splashing in shallow puddles, jacket rippling, water spraying off each stride. She cuts hard around a concrete pillar and the camera swings to follow. Cold blue ambient light with flickering fluorescents. Audio: pounding footfalls with wet slaps, heavy breathing, thunder rolling outside, no music.

Materials plus forces ("soaked windbreaker", "splashing", "rippling") is what triggers convincing physics instead of floaty motion. The hard cut around the pillar gives the handheld camera a motivated move.

Food macro with foley

16:9 · 6s · 1080p · audio on
Premium food commercial, macro lens, shallow focus. A cast-iron pan fills the frame as butter slides across it, foaming and browning; a hand lays in a scallop that sears instantly, steam and tiny oil droplets rising in backlight. Warm tungsten key light from the right. The camera creeps forward a few centimeters over the whole shot. Audio: loud close-mic sizzle, the soft clink of the scallop meeting iron, a low warm jazz bass in the background.

One continuous micro-action beats three edits at this scale. Close-mic sizzle described explicitly becomes the star of the soundtrack — food ads on Veo are half sound design.

Fast draft for prompt iteration

16:9 · 8s · 720p · Veo 3.1 Fast
Cinematic travel footage, drone-style aerial. The camera glides over a turquoise coastline at golden hour, waves breaking in slow white arcs against cliffs, a tiny sailboat leaving a wake near the frame edge. Warm low sun, long shadows on the water. Audio: wind, distant gulls, soft acoustic guitar.

Run drafts like this on Veo 3.1 Fast at 720p to test framing and palette cheaply, then rerun the exact same prompt on Veo 3.1 at 1080p for the final. The prompt transfers unchanged between the two.

Settings that matter

  • Duration

    4, 6 or 8 seconds. 8s is the sweet spot for ads and Reels — enough for a hook, a beat and a payoff. Use 4-6s for single-action shots you plan to cut together in Media Studio.

  • Resolution

    Draft at 720p, finalize at 1080p. The prompt carries over 1:1, so iterate cheap and rerun the winner.

  • Aspect ratio

    9:16 for Reels/TikTok/Stories, 16:9 for YouTube and web. Decide before writing — framing language like "selfie framing" or "wide establishing" should match the ratio.

  • Generate audio

    Leave it on and direct the sound in the prompt — native audio is half of what makes Veo output feel finished. Turn it off only when you will score the clip yourself in Media Studio.

  • Mode

    Text-to-video for from-scratch shots. Image-to-video when the opening frame must be exact — generate the still in Media Forge first, then animate it. For ads, the first-frame workflow is the reliable way to control the hook.

Do and don't

Do

  • Declare the style in the first sentence — "cinematic commercial", "UGC handheld", "documentary" — and reuse the exact wording across clips that must match.
  • Write 100-150 words in prose, 3-6 sentences: Veo rewards a full paragraph over a keyword list.
  • Give exactly one camera instruction per shot, in film language — dolly, tracking, static tripod, whip-pan.
  • Write audio cues for everything you want to hear: ambience, per-action foley, music, and dialogue in quotes with a described speaker.
  • Describe action as ordered beats — "she lifts, she pauses, she turns" — so the model paces the clip instead of compressing it.
  • Put the hook in the first two seconds when the clip is an ad — lead with the strongest action, not the context.

Don't

  • Don't write vague prompts without camera direction — that is the number one cause of flat, aimless output.
  • Don't skip the style declaration; an undeclared style is re-rolled at random on every generation.
  • Don't rely on Veo to render on-screen text, captions or logos — communicate through the native audio and add text overlays in Media Studio.
  • Don't list what to avoid — there is no negative-prompt field in Media Forge. Describe what you want instead: "an empty, spotless counter" beats "no clutter".
  • Don't stack several camera moves in one 8-second shot — one move, fully described, reads as intentional cinema; three read as chaos.
  • Don't leave dialogue unattributed — say who speaks and quote the line, or lip-sync lands on the wrong subject.

Advanced techniques

The 100-150 word treatment

Veo's controllability scales with prompt completeness up to roughly 150 words, then plateaus. The reliable shape is a short director's treatment: style, camera, subject, setting, light, two or three action beats, audio. Every element you leave unspecified is a decision the model makes for you — and remakes differently on the next run. If takes vary wildly between generations, the fix is almost always to specify more, not to regenerate more.

Directing native audio

Treat sound as a first-class layer of the prompt. Three kinds of cues work: ambience declared once ("rain on tarpaulin, distant traffic"), foley attached to visible events ("a soft clink as the scallop meets the pan"), and music described by genre and dynamics ("minimal electronic, a synth hit when the screen lights up"). Dialogue goes in quotes with the speaker described, and gets lip-synced. For multilingual ad campaigns, generate the master with native audio for timing, then decide per market whether to keep it or replace the track in Media Studio.

The first-frame workflow for ads

The highest-control pipeline on Clipwave: generate the exact opening frame with an image model in Media Forge, then feed it to Veo 3.1 image-to-video with a motion-and-audio prompt. You get a pixel-controlled hook frame, consistent branding, and a repeatable look across a whole campaign — rerun the same still with prompt variations to A/B the motion. Keep the style sentence identical across every clip in the set so cuts feel like one film.

  • Still first: lock composition, product, wardrobe and light in the image step
  • Prompt directs only motion and sound in the i2v step
  • Same style sentence across all clips of a campaign
  • Draft motion variants on Veo 3.1 Fast, finalize the winner on Veo 3.1

Which variant to use

  • Veo 3.1

    Default for finals — top cinematic quality with native audio, up to 1080p.

  • Veo 3.1 Fast

    Drafting and prompt iteration — same prompt grammar, quicker turnaround; rerun the winning prompt on Veo 3.1 for delivery.

Frequently asked

Does Veo 3.1 generate audio automatically?

Yes — audio is generated natively with the video, synchronized to the action. But it only gets specific when you direct it: write ambience, per-event sound effects, music and quoted dialogue into the prompt. The Generate Audio toggle in Media Forge turns it off entirely when you plan to score the clip yourself.

How do I make someone speak a specific line?

Describe the speaker and quote the line inside the prompt: "the barista says: 'First one is on the house.'" Veo lip-syncs the quoted dialogue to that character. Keep lines short — a few seconds of natural speech per 8-second clip.

Why does my footage look flat and generic?

Almost always a prompt with no declared style and no camera direction. Open with the style ("cinematic commercial, shot on 35mm"), give one explicit camera move, and name the light source and direction — those three lines transform the output more than any quality keyword.

Can Veo render text, logos or captions in the video?

Don't count on it — generated on-screen text is unreliable on every current video model. Let the native audio carry the message and add titles, captions and CTAs as overlays in Media Studio, where they stay sharp and editable.

Veo 3.1 or Veo 3.1 Fast — which should I use?

Both accept identical prompts, so use Fast at 720p to iterate on framing, style and audio cheaply, then rerun the final prompt on Veo 3.1 at 1080p. Drafting on the full model wastes most of its quality on takes you will throw away.

Text-to-video or image-to-video for product ads?

Image-to-video. Generate the exact opening frame with an image model in Media Forge — product, framing, light all locked — then let Veo animate it. Text-to-video is for scenes where no precise asset or hook frame has to survive.

Try Veo 3.1 in Media Forge →About Veo 3.1All models
Clipwave

Tell me what you need and grab a coffee. I'll have it ready before you finish the cup.

Product

Marketing AgentWorkflow StudioMedia ForgeTemplatesAI Models

Open in app

Workflow StudioMedia ForgeProjectsAvatars

Company

ContactPrivacy PolicyTerms of ServiceLegal Notice

© 2026 Clipwave

support@clipwave.io