Spoken dialogue is the thing Veo 3.1 does that older video models can't: it generates native synced audio — speech, sound effects, and ambience baked into the clip, with the mouth actually matching the words. These 20 prompts are all built around that one strength, grouped from simple direct-to-camera lines up to accented, emotional, two-person scenes. Paste any of them into the Gemini app, Flow, or the Gemini API, tweak the words, and generate. For the full library, start with the 35 best Veo prompts roundup.

Three rules make dialogue work. First, put the spoken words in quotation marks and name the speakerThe host says, "Welcome back." Veo reads quoted text as speech. Second, keep it to one or two short lines per clip; a generation is only 4-8 seconds, so roughly 20-30 words is the ceiling before delivery gets rushed. Third, describe the delivery (tone, accent, emotion) right before the quote, and add (no subtitles, no on-screen text) at the end so Veo doesn't burn captions onto the frame. Every prompt below follows the same Subject + Action + Setting + Camera + Lighting + Audio pattern and ends with a plain settings note like (16:9, 1080p, 8s) — duration, resolution, and aspect ratio are app settings, not inline parameters.

Advertisement

Single Speaker to Camera

The cleanest place to start. One face, a short line straight to the lens, and a mostly still camera so the lip-sync has an easy job. These are ideal for intros, ads, and hooks.

1. Direct-to-camera welcome

A friendly woman in her thirties with short curly hair looks straight into the lens and smiles as she starts talking. Bright modern home office, plants on the shelf behind her. Medium shot, static locked-off camera. Soft window light, shallow depth of field. She says, warm and upbeat, "Hey — welcome back to the channel, let's get into it." Ambient: quiet room tone. (no subtitles, no on-screen text)

Best for: YouTube and Reels intros. (16:9, 1080p, 8s)

2. Product founder pitch

A confident founder in a plain grey tee holds a small skincare bottle up beside her face and speaks to camera. Minimalist studio, seamless white backdrop. Medium close-up, very slow push-in. High-key softbox lighting, clean and even. She says, calm and sincere, "We made this because nothing else worked for us either." SFX: soft bottle tap as she sets it down. Ambient: studio silence. (no subtitles, no on-screen text)

Why it works: one line, one slow move, product held near the mouth so the eye and lip-sync land together. (9:16, 1080p, 8s)

3. News anchor cold open

A polished news anchor in a navy blazer sits at a glass desk and addresses the camera as a broadcast begins. Modern newsroom set, blurred monitors glowing behind. Medium shot, static camera. Crisp high-key studio lighting. He says, in a measured broadcast voice, "Good evening — our top story tonight is developing fast." Ambient: low newsroom hum. SFX: faint broadcast sting. (no subtitles, no on-screen text)

Best for: parody news, explainers, faux-broadcast openers. (16:9, 1080p, 8s)

4. Fitness coach hype line

A high-energy fitness coach in athletic wear points at the camera mid-rep and calls out a line. Gritty gym, weights racked behind, hard shadows. Medium close-up, subtle handheld. Low-key lighting with a bright rim light on his shoulders. He shouts, energetic and gravelly, "One more set — you've got this, let's move!" SFX: dumbbells clinking, distant treadmill. Ambient: gym music thump. (no subtitles, no on-screen text)

Why it works: emotion is stated (energetic, gravelly) so the shout reads right. (9:16, 1080p, 6s)

Two-Person Conversations

Name each speaker and let the line pass back and forth once. Keep the whole exchange to two or three short lines so it fits an 8-second clip, and frame it so both mouths are readable. For longer scenes, split across clips and use Extend. See how to prompt Veo for realistic video for pacing multi-clip scenes.

5. Café order exchange

A barista and a customer face each other across a wooden counter in a busy café. Warm morning light through the front window. Over-the-shoulder from behind the customer, static camera. Practical pendant lights, shallow depth of field. The customer says, "One oat flat white, please." The barista grins and replies, "Coming right up." SFX: espresso machine hissing, cups clinking. Ambient: quiet café murmur, soft jazz. (no subtitles, no on-screen text)

Best for: everyday, natural back-and-forth. (16:9, 1080p, 8s)

6. Office disagreement

Two colleagues stand at a glass conference table, one leaning in, the other crossing her arms. Modern open-plan office, cool afternoon light. Two-shot, static camera holding both faces. Even overhead lighting, slight window glare. He says, tense, "We can't ship it like this." She answers, flat and firm, "Then we don't ship." Ambient: distant keyboard clatter, muffled office chatter. (no subtitles, no on-screen text)

Why it works: a two-shot keeps both mouths visible for clean sync on the exchange. (16:9, 1080p, 8s)

7. Reunion on a doorstep

An older woman opens her front door to a grown son she hasn't seen in years; both freeze, then break into smiles. Suburban porch at golden hour, warm low sun. Medium two-shot, very slow push-in. Golden hour lighting, soft rim light. She whispers, trembling, "You're home." He says softly, "Hi, Mom." SFX: door creak, keys jingle. Ambient: evening crickets, distant lawnmower. (no subtitles, no on-screen text)

Best for: emotional short-film beats. (16:9, 4K, 8s)

8. Interview two-shot

A podcast host and guest sit in facing chairs, mics on booms between them, mid-conversation. Warm studio with soft neon signage behind. Wide two-shot, static camera. Low-key mood lighting, practical neon glow. The host asks, curious, "So what actually changed your mind?" The guest laughs and says, "Honestly? One bad Tuesday." Ambient: quiet studio room tone. (no subtitles, no on-screen text)

Why it works: a natural question-and-answer beat that reads as a real interview. (16:9, 1080p, 8s)

Advertisement

Character Monologues

A single character carrying a beat of drama. Give them one strong line, describe the emotion, and hold the camera close so the face does the work.

9. Detective's quiet reveal

A weathered detective in a rumpled coat leans against a rain-streaked window, turning slowly to speak. Dim precinct office at night, single desk lamp. Close-up, slow dolly in. Low-key lighting, hard shadow across half his face, practical lamp glow. He says, low and certain, "You already know who did it. You just don't want to." SFX: rain on glass, distant thunder. Ambient: buzzing fluorescent. (no subtitles, no on-screen text)

Best for: noir and thriller monologues. (16:9, 4K, 8s)

10. Rooftop confession

A young man stands at a city rooftop railing at dusk, wind moving his jacket, and speaks to someone off-frame. City skyline, blue hour sky fading to orange. Medium close-up, subtle handheld drift. Blue hour lighting, warm rim light from a doorway. He says, shaky but brave, "I should have told you the truth a long time ago." SFX: wind gusting, distant traffic. Ambient: city hum below. (no subtitles, no on-screen text)

Why it works: the stated emotion (shaky but brave) shapes a vulnerable delivery. (9:16, 4K, 8s)

11. Astronaut's last log

An astronaut floats in a dim capsule, helmet off, recording a message into a small camera. Cramped spacecraft interior, glowing instrument panels, Earth blue in the window. Close-up, static camera with a faint slow drift. Cool practical panel light, soft key on the face. She says, calm and tired, "If you're hearing this, I made it further than we thought." SFX: soft equipment beeps, air circulation hum. Ambient: low mechanical rumble. (no subtitles, no on-screen text)

Best for: sci-fi voice-log scenes. (16:9, 4K, 8s)

Narrator & Voiceover

Here the speaker is often off-screen. Attribute the line to a narrator or voiceover so Veo lays it over the visuals instead of trying to lip-sync a face. Great for docs, ads, and trailers — and a common building block in Veo prompts for YouTube Shorts.

12. Documentary narrator over landscape

A herd of elephants crosses a golden savannah as dust rises around their feet. Vast plains at sunrise, acacia trees on the horizon. Extreme wide shot, slow crane rising. Golden hour lighting, volumetric dust in the light. Narrator voiceover, warm and unhurried: "For a thousand generations, they have walked this same path." SFX: distant elephant rumble, dry grass. Ambient: morning birdsong, wind. (no subtitles, no on-screen text)

Best for: nature docs and travel films. (16:9, 4K, 8s)

13. Product explainer voiceover

A sleek wireless earbud case opens on a marble surface, the buds lifting slightly as light catches them. Clean studio, soft gradient background. Macro close-up, slow arc around the case. High-key lighting, crisp specular highlights. Narrator voiceover, bright and confident: "Twelve hours of playback in a case that fits your palm." SFX: soft magnetic click as the lid opens. Ambient: studio silence. (no subtitles, no on-screen text)

Why it works: voiceover carries the claim while the macro shot sells the product. See more ad-ready prompts. (16:9, 4K, 6s)

14. Movie-trailer narration

A lone figure walks toward a burning horizon across a cracked desert, cape whipping in the heat. Post-apocalyptic wasteland at dusk, embers drifting. Wide shot, slow dolly in behind the figure. Low-key lighting, orange backlight, heavy film grain. Deep trailer-narrator voiceover, slow and ominous: "One choice. No second chances." SFX: low wind, crackling fire. Ambient: distant rumble. (no subtitles, no on-screen text)

Best for: teaser and trailer beats. (16:9, 4K, 8s)

15. Cooking-show host aside

A cheerful chef plates a bright pasta dish, then glances up at the camera with a wink. Sunny test kitchen, herbs and utensils on the counter. Medium shot, static camera. Warm high-key lighting, shallow depth of field. She says, playful and inviting, "See? Told you it was easy." SFX: pan sizzle, spoon on ceramic. Ambient: gentle kitchen bustle. (no subtitles, no on-screen text)

Why it works: on-camera line plus SFX makes it feel like a real cooking segment. (9:16, 1080p, 6s)

Accents, Emotion & Delivery

Veo interprets plain-language voice notes. Describe the accent and emotion before the quote — a warm Scottish lilt, weary and cracking — and pair the two for the most consistent delivery.

16. Scottish tour guide

A cheerful tour guide in a wool jumper gestures toward an old stone castle on a misty hill. Scottish highlands, overcast morning, heather in the foreground. Medium shot, slow tracking beside him. Soft diffused daylight, moody grey sky. He says, in a warm Scottish accent, "Now this old fortress has stood eight hundred years — and a few ghosts too." SFX: wind over grass. Ambient: distant bagpipes, birdcall. (no subtitles, no on-screen text)

Best for: travel and accent tests. (16:9, 4K, 8s)

17. Southern US storyteller

An old man rocks slowly on a wooden porch, a sweet tea beside him, leaning in to tell a story. Rural front porch at dusk, warm porch light on. Close-up, static camera. Golden practical porch light, soft warm shadows. He says, in a slow Southern drawl, "My granddaddy built this whole place with his two bare hands." SFX: chair creak, cicadas. Ambient: warm evening hum. (no subtitles, no on-screen text)

Why it works: the drawl note plus a slow rock give the delivery its unhurried rhythm. (16:9, 4K, 8s)

18. London cabbie banter

A cheerful black-cab driver glances in the rear-view mirror to chat with a passenger. London streets at night, wet roads reflecting neon and traffic lights. Close-up over the seat, subtle handheld. Low-key lighting, practical dashboard glow, streaks of passing light. He says, in a chirpy London accent, "Where to, love? Traffic's murder tonight, mind." SFX: engine idle, indicator tick. Ambient: city traffic, light rain. (no subtitles, no on-screen text)

Best for: character voice and city vibe. (9:16, 1080p, 8s)

19. Tearful goodbye

A woman stands on a train platform holding a suitcase, eyes brimming, speaking to someone off-frame as a whistle sounds. Old railway station, cold morning light, steam drifting. Close-up, very slow push-in. Cool diffused light, soft haze. She says, voice cracking, "I don't want to say goodbye — so I won't." SFX: train whistle, hydraulic hiss. Ambient: distant announcements, footsteps on tile. (no subtitles, no on-screen text)

Why it works: "voice cracking" tells Veo to break the delivery with emotion. (16:9, 4K, 8s)

20. Furious argument in the rain

A man and a woman stand inches apart in a downpour, both soaked, shouting over the storm. Empty city street at night, rain hammering, neon puddles. Medium two-shot, subtle handheld. Low-key lighting, harsh neon rim light, heavy rain. She yells, furious, "You never even asked me!" He shouts back, raw, "Because I already knew the answer!" SFX: pouring rain, distant thunder. Ambient: storm roar. (no subtitles, no on-screen text)

Best for: high-drama scenes. Keep both lines short so the sync survives the fast delivery. (16:9, 4K, 8s)

Want to pair these with silent B-roll or product shots? The full 35 best Veo prompts roundup covers every use case, and how to prompt Veo for realistic video goes deeper on why some lip-sync looks flawless and some drifts.

Frequently Asked Questions

How do I write dialogue in a Veo prompt?

Name the speaker, then put the exact spoken words in quotation marks — for example: The barista says, "One oat flat white, coming up." Veo 3.1 reads the quoted text as speech and generates lip-synced audio for it. Keep it to one or two short lines per clip, since a clip is only 4-8 seconds long, and describe the tone or emotion just before the quote so delivery matches.

Why do captions keep appearing on my clip, and how do I stop them?

Veo 3.1 sometimes burns subtitles onto the frame when it detects spoken dialogue. Add (no subtitles, no on-screen text) at the end of the prompt to suppress them. If a caption still slips through, regenerate — the phrasing works most of the time but isn't a hard guarantee, so a re-roll usually clears it.

How accurate is Veo 3.1's lip-sync?

Lip-sync is strongest on medium shots and close-ups where the mouth is clearly visible and the line is short. Give the character one or two short sentences, keep the camera fairly still, and avoid extreme angles or fast motion that hide the mouth. Long, rushed lines are where sync drifts most, so trim the words before you blame the model.

Can Veo handle two people talking to each other?

Yes. Name each speaker and give them a short line in sequence — for example: Maya says, "You're late." Then Sam replies, "Traffic." Keep the total to two or three short lines so it fits an 8-second clip, and use an over-the-shoulder or two-shot so both mouths are readable. For longer exchanges, split the conversation across clips and use Extend.

Can I control the character's accent or voice?

Describe the voice in plain words before the quote — for example a warm Scottish accent, a deep gravelly voice, a bright American radio-announcer tone. Veo interprets these descriptions and shapes the delivery accordingly. Pair the accent note with an emotion (calm, excited, weary) for the most consistent result.

How long a line of dialogue fits in an 8-second clip?

Roughly 20-30 spoken words, or one to two short sentences, fits comfortably in 8 seconds at a natural pace. Anything longer gets rushed and the lip-sync suffers. If your line is longer, cut it down or split it across two clips and stitch them with scene extension.

Do I need a reference image to keep the same character talking across clips?

For a multi-clip conversation, use Ingredients to Video: upload one reference image of the character and reuse it across each generation to keep the face, wardrobe, and voice consistent. Frames to Video also helps — end one clip and seed the next with its last frame so the cut feels continuous.

Should I set duration and aspect ratio inside the prompt?

No. Duration (4, 6, or 8 seconds), resolution (720p, 1080p, or 4K), and aspect ratio (16:9 or 9:16) are app and API settings in the Gemini app, Flow, or Vertex AI — not inline parameters. The (16:9, 1080p, 8s) note at the end of each prompt is a reminder of which settings to pick, not text Veo reads.

Advertisement