PromptFork

Scene prose. Temporal beats. Weight.

Veo prompts: write the scene Google’s video model can actually hold

Veo is not a keyword salad generator. It thrives on readable scene description: where we are, what continuous action unfolds, how the camera spends the seconds, what materials should obey, and what must never appear. That is a different dialect from Sora’s tighter shot tags — which is why this page exists separately. Below is a free Veo Shot Builder, then the system for debugging almost-right clips.

Veo Shot BuilderFree · instant

3. Camera

4. Duration & beats

5. Look (Veo style)

6. Environment

7. Time of day

Describe a subject to start.

Composed in your browser — paste into Veo when ready. Here is why the sections look the way they do.

Why Veo wants scene prose (and why that is good news)

If you come from image models, you may still type comma-separated aesthetic tags and hope motion appears. Sometimes it does. Often you get a beautiful mess: “cinematic, 8k, epic, masterpiece, slow mo, drone, emotional” with no verb strong enough to organize time. Veo is happier with sentences that a human DP could follow: In a mid-morning loft kitchen, hot water spirals into a pour-over; steam rises; the camera dollies in over eight seconds; liquid physics stay coherent; no text. The second prompt allocates attention across space and time.

Temporal structure is the hidden superpower. Humans experience clips as beginnings, middles, and ends even when they cannot name the beats. When you specify establish → develop → settle, you reduce aimless motion and unfinished actions. When you match beat count to duration, you stop asking a four-second clip to deliver a short film.

Physics language is the other half. Cloth, liquid, hair, foliage, rain — secondary motion sells reality. Asking for “weighty continuous motion” and “preserved materials” is not pretension; it is how you keep glass from melting and faces from drifting. The builder writes those clauses so you do not forget them under deadline.

Anatomy of a Veo prompt

Scene

Place + subject + action in prose. Lead with environment and time when light matters; lead with the hero object when the product is the story. Use distinctive materials and colors. Keep action continuous — one physically plausible chain, not a montage hidden in commas.

Camera and lensing

Dolly in/out, pan follow, tilt up, aerial drift, over-shoulder, POV, wide locked. One idea. Add “no random reframes” so the model does not invent a second camera operator mid-clip.

Temporal structure

This section is Veo-native discipline. Short clips: one or two beats. Eight to twelve seconds: establish, develop, settle (or aftermath on a detail). You are budgeting attention the way an editor budgets cuts — except you are doing it before generation.

Look and rendering

Photoreal, prestige drama, nature doc, product ad, grounded sci-fi, period, 3D animation, street verité — pick a family. Conflicting families create style thrash. Mention secondary motion relevant to the scene so the look is not only color grade cosplay.

Physics, continuity, and do-not-include

Preserve identity, scale, materials, light direction. Forbid text, watermarks, sudden cuts, morphs, impossible physics. Domain-specific bans (liquid clipping, warped reflections) belong here after you have seen the failure once.

How to use the Veo Shot Builder

Subject and action first. Camera and duration next — duration chooses the beat template. Look after that so style cannot contradict the story. Environment and time of day lock the world. Generate, paste into Veo, and when something is off, edit the matching section only. Supercharge routes the brief through Studio if you want tighter prose before you spend a generation.

Five Veo prompt recipes you can steal

1

Product pour (photoreal commercial)

Generate a 8s video clip for Google Veo.

Scene:
In minimal loft interior at mid-morning, we see a ceramic pour-over kettle and glass carafe on a maple kitchen counter. Hot water streams in a slow spiral into the filter; steam rises; the carafe fills with amber coffee.

Camera & lensing:
Smooth dolly-in that gradually tightens on the subject. Maintain consistent framing logic for the full duration; no random reframes.

Temporal structure:
Beat 1 establish → Beat 2 develop action → Beat 3 settle on a clear final frame.

Look & rendering:
Premium product commercial: immaculate surfaces, hero lighting, fluid but purposeful motion, brand-safe clarity. Motion should feel continuous and weighty — liquid and steam show natural secondary motion.

Physics & continuity:
Preserve subject identity, scale, and material properties from the first frame to the last. Lighting direction stays consistent.

Do not include:
on-screen text, logos, shaky zoom, liquid clipping through glass, warped reflections

Why it works — Liquid + glass is a physics stress test; the prompt names secondary motion and forbids clipping.

2

Nature doc patience

Generate a 12s video clip for Google Veo.

Scene:
In quiet forest clearing at dawn, we see a heron standing in shallow water among reeds. It waits, then spears once and lifts a small fish, water droplets scattering.

Camera & lensing:
Wide locked-off establishing frame. Maintain consistent framing logic for the full duration.

Temporal structure:
Beat 1 establish environment → Beat 2 subject action peaks → Beat 3 aftermath / reaction held on detail.

Look & rendering:
BBC-style nature documentary realism: patient framing, authentic animal behavior, true-to-life color.

Physics & continuity:
Preserve subject identity and scale; water and reeds show natural secondary motion.

Do not include:
on-screen text, watermarks, sudden cuts, cartoon animals, impossible physics

Why it works — Locked wide + long beats match nature-doc timing instead of music-video chaos.

3

Drama over-shoulder

Generate a 8s video clip for Google Veo.

Scene:
In rain-slick alley at deep night, we see two people under a single practical lamp, coats wet, breath visible. The nearer person hands over a small sealed envelope; neither speaks; hands linger a moment.

Camera & lensing:
Over-the-shoulder framing with shallow depth of field. Maintain consistent framing logic.

Temporal structure:
Beat 1 establish → Beat 2 develop action → Beat 3 settle on a clear final frame.

Look & rendering:
Prestige drama cinematography: controlled contrast, expressive color grade, intentional negative space, theatrical blocking.

Physics & continuity:
Preserve faces and wardrobe identity; rain and fabric show weighty secondary motion; light direction stays consistent.

Do not include:
on-screen text, jump cuts, identity morphing, exaggerated VFX

Why it works — Story is a single prop handoff — Veo can hold faces if you forbid morph and multi-beat plots.

4

Aerial establishing

Generate a 6s video clip for Google Veo.

Scene:
In coastal cliffs at golden hour, we see a narrow road hugging the cliff edge above turquoise water. A single vintage car moves along the road, tiny against the landscape.

Camera & lensing:
High aerial drone perspective with gentle forward drift.

Temporal structure:
Beat 1 establish → Beat 2 complete one continuous action.

Look & rendering:
Photoreal live-action footage with accurate materials and physically plausible light transport.

Physics & continuity:
Preserve scale relationships; no teleporting car; consistent sun direction.

Do not include:
UI overlays, map pins, text, impossible drone whip moves

Why it works — Aerial shots need scale anchors and restrained camera drift more than adjective stacks.

5

Sci-fi grounded detail

Generate a 8s video clip for Google Veo.

Scene:
In dense city street at blue hour, we see a courier in a worn jacket adjusting a compact holographic wrist map that flickers softly. Pedestrians pass; the courier starts walking with purpose.

Camera & lensing:
Horizontal pan that follows the subject’s path.

Temporal structure:
Beat 1 establish → Beat 2 develop action → Beat 3 settle on a clear final frame.

Look & rendering:
Grounded science-fiction production design: believable tech details, volumetric atmosphere, restrained VFX integration. Weight secondary motion on coat and crowd.

Physics & continuity:
Keep the same courier identity; hologram stays attached to the wrist; no style thrash into pure cartoon.

Do not include:
on-screen title cards, chaotic laser battles, morphing face, low-res blur

Why it works — “Grounded sci-fi” plus restrained VFX stops the clip from becoming a trailer mashup.

When Veo drifts: symptom → section to edit

Action feels rushed or incomplete

Increase duration or simplify action; check temporal beats match the window.

Identity morphs or wardrobe drifts

Strengthen physics/continuity; remove multi-person chaos; forbid morphing explicitly.

Camera movement fights the story

Pick one camera phrase; for performance, try wide locked or over-shoulder only.

Wrong genre vibe

Rewrite Look section to a single family; delete conflicting medium words.

Environment changes mid-clip

Lock place and time of day in Scene; ban day-to-night transitions in one shot.

Liquids/cloth look fake

Call out secondary motion and weight; add domain-specific do-not-include failures.

Text or UI junk appears

Do-not-include on-screen text, logos, watermarks, map pins, title cards.

Looks like a trailer mashup

One action only; remove “epic battle then escape then…” chains; shorten ambition.

Veo vs Sora vs other PromptFork tools

NeedDialectGo to
Google Veo clipScene prose + beats + physics sectionsThis page
OpenAI Sora clipTight shot lines + camera tags + avoidSora prompts
Still frames firstSubject/style/light/lens/negativeAI image prompts
Research before conceptingSources + recency + depthPerplexity prompts
App demo scaffoldingVision/pages/acceptanceLovable prompts
Voiceover script draftRole/task/format from a goalChatGPT generator

Cross-link deliberately: concept in Perplexity, board with image prompts, motion in Veo or Sora, script with the ChatGPT Prompt Generator. Same network, different grammars.

Keep a Veo library that encodes your taste

Your best “do not include” lines are hard-won. So are your temporal templates and product-ad look paragraphs. Fork them. The subject changes; the craft stays. That is how a personal Veo style emerges without retyping physics essays every Monday.

Find

Browse shot briefs by look family — product, nature doc, drama — not by random chat history.

Copy

Reuse section scaffolding; only rewrite Scene subject and action.

Fork

Save the version that finally fixed liquid physics or face stability.

For model-agnostic prompt skill, keep FORGE in your rotation. For weak chat prompts that need rewrites rather than greenfield generation, use the ChatGPT optimizer. Explore and top lists on PromptFork surface what the community already stress-tested.

A sane Veo production loop

Lock the scene sentence before you touch style adjectives. Generate short, readable clips before long heroic ones. When stakeholders say “more premium,” map the note to look and light, not to five new plot beats. When they say “more real,” map to physics and secondary motion. Keep audio ambitions separate when possible — picture first, score and VO after — unless your pipeline truly generates synced sound you trust.

Version the prompt alongside the clip filename. Future you will not remember which continuity clause saved the hero take. PromptFork is built for that memory: find, copy, fork, studio refine, ship.

If you are choosing between Veo and Sora for a project, choose based on ecosystem and the dialect you want to maintain in your library — then stay consistent inside a sequence so shots feel related. Dual-library is fine; dual-dialect inside one prompt is not.

Treat temporal beats like a shot list inside a single generation. Before you click generate, say out loud what happens in the first third, middle third, and final third of the clip. If you cannot, the model cannot either. Empty middles produce wandering cameras; overloaded openings produce unfinished actions. Writing the three beats in the prompt is simply making that spoken plan binding.

Materials deserve explicit nouns. “Coffee” is weaker than “amber coffee filling a clear glass carafe.” “Car” is weaker than “vintage teal coupe with wet chrome bumpers.” Veo’s photoreal modes use those material anchors to keep reflections and color stable. When a clip goes muddy, the fix is often not more adjectives — it is better nouns in the scene paragraph.

For product work, build a “hero kit” prompt: locked studio or loft environment, locked time of day, locked look section, and a do-not-include list tuned to your category (labels warping, liquid clipping, logo invention). Only the product description and the action line change between SKUs. That kit is how a catalog of clips stays on-brand without a human DP on every still-to-motion job.

For narrative work, protect faces. If identity stability is the product, simplify backgrounds, reduce crowd density, and forbid morphing in the same breath as you describe the person. Over-shoulder framing can help because it shows less of the secondary face. POV trades face risk for motion sickness risk — choose consciously. Every framing choice is a risk budget.

Close the loop with review criteria written before generation: “action completes,” “no text,” “light direction stable,” “same jacket start to end.” Score takes against that list instead of against a vague hope that it feels cinematic. Cinematic is a result of criteria you can name. Named criteria are also what you encode back into the next fork of the prompt when a take finally clears the bar.

Surgical section edits beat full rewrites

Veo’s sectioned structure is a gift for iteration. When a take fails, edit only the section that failed. Wrong motion? Rewrite temporal beats, leave subject alone. Wrong mood? Adjust light and atmosphere, leave camera path alone. Full rewrites reintroduce problems you already solved. Treat each section like a git file: small diffs, clear commit notes, fork when the product needs a permanent branch.

Keep a “known good” base prompt per product line with locked wardrobe, palette, and lens language. Campaign variants only swap subject action and location. That is how a catalog stays on-brand when five freelancers generate in parallel. The shot builder on this page is the fastest way to draft the first structured version; the library is how that draft becomes institutional memory instead of a one-off chat win.

Questions people ask about Veo prompts

What makes a good Veo prompt?+

A good Veo prompt reads like a precise scene description for a camera crew: where we are, what we see, what continuous action unfolds, how the camera moves, how time is structured across the clip, what the look should feel like, and what must not appear. Veo responds well to fuller prose and explicit temporal beats — establish, develop, settle — plus physics and continuity language that keep materials and identity stable.

How are Veo prompts different from Sora prompts?+

Both need subject, motion, and camera, but the dialect differs. Sora prompts on PromptFork lean tighter film-tag lines. Veo prompts lean scene prose, labeled sections (Scene, Camera, Temporal structure, Look, Physics), and weighty secondary motion. Reusing a Sora tag salad in Veo often under-specifies time structure; pasting a Veo essay into Sora can overstuff. Use this builder for Veo and the Sora Shot Builder for Sora.

How do temporal beats improve Veo results?+

Temporal beats tell the model how to spend the seconds: establish environment, develop action, settle on a final frame. Without beats, motion can feel aimless or front-loaded. Matching beat count to duration (shorter clips get fewer beats) keeps physics readable and endings intentional.

Should Veo prompts emphasize photoreal physics?+

When you want live-action plausibility, yes. Call out weighty motion, secondary motion (cloth, liquid, hair, foliage), consistent light direction, and preserved materials. If you want stylized 3D or animation, say that in the look section instead — but still demand coherent stylization so the clip does not thrash styles mid-shot.

How detailed should the environment and time of day be?+

Enough to lock palette and light logic: “minimal loft interior at mid-morning” is more useful than “nice room.” Environment plus time of day prevents the model from inventing a conflicting setting halfway through. If place is the story, lean harder; if product is the story, keep environment simple and controlled.

Why does my Veo clip ignore part of the prompt?+

Usually overcrowding: multiple actions, competing camera moves, or conflicting looks. Compress to one continuous action, one camera idea, and one look family. Put must-nots in a short do-not-include list. If a detail is critical, put it in the opening scene sentence, not buried at the end of a paragraph.

Can I use these prompts for other Google video tools?+

The scene-prose + beats + physics pattern travels well across Google’s video generation experiences and many text-to-video systems. Always verify current UI limits (duration, aspect, audio). Keep a library of Veo-native prompts and retune only the dialect edges when tools change names.

Does the free Veo builder send data to Google?+

No. Prompt composition runs locally in your browser. You copy the result into Veo yourself. Supercharge opens PromptFork Studio to refine the prompt text under Studio’s free limits — separate from Google’s services.

What belongs in “Do not include” for Veo?+

On-screen text, watermarks, logos, sudden cuts, identity morphing, impossible physics, low-res blur — plus any domain-specific artifacts you have seen (liquid clipping through glass, warped reflections). Keep the list short and failure-driven.

What does forking a Veo prompt mean on PromptFork?+

Forking copies a proven Veo scene brief into your library so you can change the subject while keeping temporal structure, camera language, and physics rules that already held up. It is how you build a reusable directing style for Google’s video model.

Give Veo a scene, not a mood board dump.

Build a sectioned Veo prompt with beats and physics — or fork one that already holds continuity under pressure.