How to Prompt Veo for Realistic Video (2026)

Realistic Veo 3.1 video comes from writing like a director's shot brief, not a wish list. Nail one subject, one action, one camera move, specific light, and matched native audio, and the clip reads as filmed rather than generated. This guide gives you the exact 6-part formula, then 10 finished prompts you can paste into the Gemini app, Flow, or the Gemini API right now. For a broader library once you have the pattern down, see the 35 best Veo prompts.

The 6-part Veo formula

Every strong Veo shot is one flowing paragraph built from six beats, in this order, followed by the audio lines. Write each beat once and be concrete — Veo rewards specificity and punishes vagueness.

Assembled, a shot reads like this: "A weathered fisherman in a yellow oilskin hauls a rope over the gunwale of a small boat. Open sea at dawn, grey swell. Medium shot, slow dolly in. Overcast soft light, 35mm, film grain. SFX: rope creaking, water slapping the hull. Ambient: gulls, distant wind." That is the whole method — the rest of this guide sharpens each beat. Grab ready vocabulary from the Veo prompt cheat sheet when you need shot sizes and moves fast.

Writing native audio (the thing that sells realism)

Veo 3.1's headline feature is native synced audio: it generates dialogue, sound effects, and ambient sound together with the picture. Silent or generically-scored clips read as AI instantly, so describe sound on every prompt. Use three tools:

• Dialogue in quotes. Attribute the line to a subject and keep it short — a clip is only 4-8 seconds, so one or two lines: The barista says, "One oat flat white, coming up."
• SFX, spelled out. Name the specific sounds the action makes: SFX: espresso machine hissing, cups clinking.
• Ambient bed. Set the background layer: Ambient: quiet café murmur, soft jazz.
• No subtitles. Veo sometimes burns in captions when it hears dialogue. End the prompt with (no subtitles, no on-screen text) to suppress them.

Match the audio to the shot size, too. A close-up implies intimate, present sound; a wide establishing shot implies distant, reverberant ambience. When dialogue and SFX line up with what the camera shows, the brain accepts the clip as real. See Veo prompts for talking characters for lip-sync-friendly patterns.

Camera & lighting language that avoids the AI look

The "AI look" usually comes from motion that no real operator would produce and light with no source. Fix both with plain film vocabulary.

Camera: pick one shot size and one move. Motivated, gentle moves read as real — a slow dolly in for emphasis, a tracking shot beside a walking subject, a handheld feel for documentary energy, or a static locked-off frame that lets the action breathe. Avoid drifting, floating moves with no anchor; they are the classic tell.

Lighting: name a real source and quality. Golden hour gives warm low sun and long shadows; overcast soft light flatters faces; practical lights (lamps, neon, candles) ground a night scene; low-key builds mood with hard shadows. Add a lens (35mm documentary, 85mm portrait) and optional shallow depth of field or film grain for texture. "Cinematic lighting" alone is too vague — Veo needs a direction and a quality to work with. Pull terms from the cheat sheet tables when you are stuck.

Consistency across shots (Ingredients, Frames, Extend)

Realism breaks the moment a character's face or a product changes between shots. Veo 3.1 has three features for continuity, and they belong in your workflow, not your prompt text:

• Ingredients to Video: upload up to about three reference images — a character, a product, a style board — so Veo keeps them consistent across separate generations. Essential for branded product shots and recurring characters.
• Frames to Video: supply a first and/or last frame to lock exactly how a shot starts and ends. Great for matching an edit or hitting a precise product hero pose.
• Extend: continue a sequence past 8 seconds by seeding the next clip with the last frames of the previous one. Build longer scenes clip by clip instead of asking for one impossible long take.

Keep each prompt describing a single shot, and use these features to chain shots together. For product accuracy specifically, pair Ingredients to Video with the recipes in Veo prompts for product & ad video.

10 copy-paste example prompts

Each prompt below is a complete 6-part shot with a described audio bed, a Why it works line, and a plain settings note. Duration, resolution, and aspect ratio are app settings — the note tells you what to pick, not text to paste into the prompt.

1. Café barista (dialogue)

A young barista with rolled-up sleeves steams milk behind a busy espresso bar, glancing up at a customer with a quick smile. Small specialty café, mid-morning light through a front window. Medium shot, slow dolly in. Warm practical lights, shallow depth of field, 35mm. The barista says, "One oat flat white, coming up." SFX: espresso machine hissing, cups clinking. Ambient: quiet café murmur, soft jazz. (no subtitles, no on-screen text)

Why it works: one line of dialogue, one gentle move, and synced SFX that match exactly what the hands are doing. (16:9, 1080p, 8s)

2. Rain-slicked city walk (mood)

A woman in a long charcoal coat walks briskly down a wet city sidewalk at night, collar up against the drizzle. Downtown street, neon signage reflected in puddles. Wide shot, tracking alongside her at a steady pace. Low-key lighting, practical neon, film grain. SFX: footsteps on wet pavement, distant car horn. Ambient: light rain, muffled traffic hum.

Why it works: a single motivated tracking move and reflective practicals give real depth without any camera drift. (9:16, 1080p, 8s)

3. Product hero — wireless earbuds (ad)

A matte-black earbud case rests on a brushed-concrete slab, lid slowly opening to reveal the earbuds and a soft glow inside. Minimal studio set. Extreme close-up, static locked-off camera. Soft top light with a subtle rim, shallow depth of field, 85mm. SFX: a crisp magnetic click as the lid opens. Ambient: near silence, faint studio room tone. (no subtitles, no on-screen text)

Why it works: a locked-off frame plus one precise SFX cue keeps focus on the product; pair with Ingredients to Video for exact branding. (16:9, 4K, 6s)

4. Mountain establishing shot (nature)

A lone hiker crests a rocky ridgeline as morning fog drifts through the valley below. Alpine range at sunrise. Extreme wide shot, slow crane rising to reveal the full valley. Golden hour light, volumetric haze through the fog, 35mm. SFX: gusting wind, gravel underfoot. Ambient: distant birdsong, open-air stillness.

Why it works: the crane is the only move, so the reveal stays smooth and the scale reads as filmed. (16:9, 4K, 8s)

5. Kitchen how-to (creator / Shorts)

A home cook cracks an egg one-handed into a steel bowl, then whisks briskly as flour dusts the counter. Bright home kitchen, midday. Close-up over the bowl, static camera. High-key soft daylight, shallow depth of field. The cook says, "See how fast that comes together?" SFX: shell cracking, whisk clinking against steel. Ambient: quiet kitchen, faint radio. (no subtitles, no on-screen text)

Why it works: one clear action, tight framing, and diegetic sound make a vertical clip feel like a real tutorial. (9:16, 1080p, 8s)

6. Vintage motorcycle reveal (action)

A vintage motorcycle rolls out of a dark tunnel onto an open desert highway at dusk, rider silhouetted against the light. Empty road, blue-hour sky fading to orange. Full shot, dolly out as the bike accelerates toward camera. Backlight from the low sun, film grain. SFX: engine growl building, tyres over grit. Ambient: dry desert wind.

Why it works: backlight and a single dolly-out sell the reveal; the audio ramp matches the acceleration. (16:9, 4K, 8s)

7. Interview close-up (talking head)

A middle-aged woodworker in a canvas apron speaks directly to camera, sawdust on her forearms, workshop tools blurred behind her. Small workshop, warm afternoon light. Close-up, static camera. Soft window key light with practical work-lamp fill, 50mm, shallow depth of field. She says, "I've been building tables for twenty years." SFX: faint saw hum in the background. Ambient: quiet workshop room tone. (no subtitles, no on-screen text)

Why it works: a still frame and one short line keep lip-sync clean and the delivery natural. (16:9, 1080p, 6s)

8. Skincare pour (beauty ad)

A glass dropper releases a single droplet of clear serum onto the back of a hand, the drop spreading slowly across the skin. Clean studio surface, soft pastel backdrop. Extreme close-up, slow push-in. Diffused softbox light, subtle rim, 100mm macro, shallow depth of field. SFX: a soft liquid tap as the drop lands. Ambient: calm studio silence. (no subtitles, no on-screen text)

Why it works: a macro push-in and a single delicate SFX give a premium, tactile feel — a natural for product ads. (9:16, 4K, 6s)

9. Coastal B-roll (cinematic)

Waves roll over black volcanic rocks as sea spray catches the low sun. Rugged coastline, late golden hour. Wide shot, slow handheld drift. Golden hour backlight, anamorphic feel, film grain. SFX: waves crashing, spray hissing over rock. Ambient: steady ocean roar, distant gulls.

Why it works: a subtle handheld drift adds life without breaking realism — ideal cutaway footage for cinematic B-roll. (16:9, 4K, 8s)

10. Street-food night market (atmosphere)

A vendor tosses noodles in a blazing wok, flames leaping as steam rises over a crowded night market stall. Busy market lane after dark. Medium shot, static camera. Warm practical lights and neon signage, low-key contrast, 35mm. The vendor calls out, "Two minutes!" SFX: wok sizzling, flame whoosh, ladle scraping. Ambient: market crowd chatter, distant music. (no subtitles, no on-screen text)

Why it works: layered practicals plus dense, matched sound make a static frame feel alive and unmistakably real. (9:16, 1080p, 8s)

Common mistakes to avoid

Most unrealistic Veo clips fail on the same handful of errors. Check every prompt against this list before you generate.

• Stacking camera moves. A pan that also dollies and cranes produces jittery, unstable motion. Use exactly one move per shot.
• Too many actions. Three verbs in one clip (she waves, sits, and pours) fights the 4-8 second window. Pick one action; use Veo prompt templates and Extend to chain more.
• Vague lighting. "Cinematic lighting" gives Veo nothing to aim for. Name a source and quality — golden hour, practical lamps, overcast soft light.
• Over-long dialogue. A paragraph of speech won't fit and won't lip-sync. Keep it to one or two short lines in quotes.
• Writing duration in the prompt. "10 seconds" in the text is ignored. Duration (4, 6, or 8s), resolution, and aspect ratio are app or API settings — note them separately, like (16:9, 1080p, 8s).
• Skipping the audio lines. Silent prompts waste Veo's biggest advantage. Always add an SFX line and an Ambient line, and (no subtitles, no on-screen text) if you don't want captions.

Fix these six and your hit rate on realistic takes climbs sharply. When you're ready for a bigger set of finished shots, work through the 35 best Veo prompts.