Veo 3.1 Prompts: Dialogue, Sound and Cinematic Shots

Veo 3.1 writes a whole scene from one prompt: picture, camera move, dialogue and sound effects, all landing in sync. The trick is that it reads your prompt like a mini script, not a mood board, so vague adjectives do less than a clear line of dialogue or a named sound. Below is the syntax that actually works, ten prompts you can copy straight into the box, and what each clip costs.

What Veo 3.1 actually does with a prompt

Veo 3.1 is Google DeepMind's video model, and the headline change from Veo 3 is audio quality. Veo 3.1 excels at generating realistic, synchronized sound, from multi-person conversations to precisely timed sound effects, all guided by the prompt. That means the model isn't bolting generic music onto a clip afterward — it's deciding, from your words, what a door should sound like when it slams and when a character's mouth should move.

The other practical limits to know before you write anything: clips run 4, 6, or 8 seconds, in 16:9 or 9:16, with audio that can be turned on or off. And the model generation itself runs at 1080p at 24 fps, with more lifelike textures, lighting, and smoother motion with fewer visual artifacts than the original Veo 3.

A Veo 3.1 prompt works best as a shot description plus a sound description, written as two connected sentences rather than a pile of keywords.

The anatomy of a working prompt

Write in this order and you'll waste fewer generations:

1. Camera and shot type — "handheld medium shot," "slow dolly in," "static wide shot."
2. Subject and action — who's in frame and what they're physically doing.
3. Setting and light — where this is, time of day, light quality.
4. Dialogue — exact words, in quotes, with a tone.
5. Sound — effects and ambience named specifically.
6. Style — "shot on 35mm," "documentary," "anime."

How to write dialogue so it lip-syncs

Don't describe speech indirectly ("he says something encouraging"). Veo 3.1 needs the literal line. The convention that reliably works across prompting guides is to name the speaker, then put the exact words you want the character to speak inside quotation marks, preceded by a colon, with a tone word right after. For example: the barista slides the cup across and says: "Careful, that's hot," with a warm smile. Keep it to one or two short lines per clip — an 8-second clip only has room for a sentence or two before the pacing feels rushed.

If two people are talking, give both lines and separate the turns clearly ("the first man says: '...' then the second man replies: '...'"). Veo 3.1 handles short back-and-forth exchanges, but cramming a long conversation into one clip usually costs you lip-sync accuracy on the second line.

How to write sound without describing music that never plays

Treat sound effects as objects in the scene, not mood words. "Tense atmosphere" gives the model nothing to render; "a low ticking clock and distant sirens" does. Name:

• Diegetic sound effects tied to visible actions — footsteps on gravel, a glass set down, rain on a window.
• Ambience that sets the room — café chatter, wind, a crowd.
• Dialogue tone, separately from the words themselves — flat, urgent, amused.

Skip instructions like "add epic orchestral score" unless the shot is clearly cinematic — Veo 3.1 will do it, but generic scoring is the fastest way to get a clip that feels like stock footage.

Do it in Plurel

Veo 3.1 and Veo 3.1 Fast are both in Plurel's video tool, on the same credit balance as every other model.

1. Open plurel.tech and sign in — it's free, no card required.
2. Under the composer, click Video, or go straight to https://plurel.tech/?model=veo3 for full Veo 3.1, or https://plurel.tech/?model=veo3-fast for the Fast tier.
3. Paste your prompt — camera, action, dialogue in quotes, named sound effects — into the box.
4. Check the credit cost shown before you confirm: Veo 3.1 Fast starts at 20 credits, full Veo 3.1 starts at 85 credits, both for an 8-second 1080p clip with audio.
5. Generate. If the clip almost works, use Extend (https://plurel.tech/?model=video-extend) to continue it, or just reply in the chat ("same scene, but the rain gets heavier") — the follow-up edits what was just made rather than starting over.
6. To animate a still photo with Veo 3.1 instead of writing from scratch, use Animate (https://plurel.tech/?model=animate) and add your dialogue and sound cues to the motion prompt.

If you'd rather describe the scene in plain language and have a model handle the shot list, GPT-6 Astra, Claude Opus 5.5 or Claude Sonnet 5.5 can turn a rough idea into a structured Veo 3.1 prompt and show you the credit cost before anything renders — open any of them from the Chat tab.

Ten Veo 3.1 prompts you can copy

Static wide shot, a lighthouse keeper on a rocky cliff at dusk, waves crashing below. He turns to camera and says: "Storm's coming in fast," with quiet urgency. Sound: heavy surf, distant foghorn, wind gusts. Style: moody, desaturated, 35mm film.

Coastal drama with dialogue and weather sound — run on Veo 3.1 (veo3).

Handheld medium shot, a street food vendor flipping noodles in a wok over high flame. Sound: sizzling oil, clanging metal, distant scooter horns, market chatter. No dialogue. Style: documentary, warm tungsten light.

Pure ambience and SFX test with no lip-sync risk — run on Veo 3.1 Fast (veo3-fast).

Slow dolly in on two coworkers at a desk. The first says: "Did you see the numbers?" flatly. The second replies: "Yeah. We're over budget," with a tired sigh. Sound: office hum, a keyboard clicking, muffled printer. Style: flat corporate lighting, 16:9.

Two-person dialogue exchange — run on Veo 3.1 (veo3) for the stronger lip-sync.

Static shot of a dog sitting by a front door, ears perking up. Sound: keys jingling off-screen, a lock turning, the door creaking open. No dialogue, no music. Style: naturalistic daylight.

Sound-effect timing test tied to an unseen action — run on Veo 3.1 Fast (veo3-fast).

Tracking shot following a chef plating a dish in a busy kitchen. She says: "Table nine, go now," sharply, without looking up. Sound: clattering pans, a timer beeping, overlapping kitchen shouts in the background. Style: fast-paced, handheld, warm kitchen light.

Dialogue layered over busy background sound — run on Veo 3.1 (veo3).

Wide static shot of an empty subway platform at night, a single train arriving. Sound: screeching brakes, a hiss of doors opening, an automated announcement: "Doors closing." Style: cold fluorescent light, slight grain.

Announcement-style dialogue plus mechanical SFX — run on Veo 3.1 Fast (veo3-fast).

Close-up on an old man's hands carving wood on a porch. He mutters to himself: "Almost there," with a soft chuckle. Sound: wood shavings falling, birds in the distance, a creaking porch swing. Style: golden hour, shallow depth of field.

Quiet single-line dialogue with layered ambience — run on Veo 3.1 (veo3).

Medium shot, a news anchor at a desk looking into camera. She says: "Good evening, and welcome back," with practiced composure. Sound: faint studio hum, no music. Style: clean broadcast lighting, 16:9.

Straight-to-camera delivery for talking-head style clips — run on Veo 3.1 Fast (veo3-fast).

Handheld shot, two kids running through a sprinkler in a backyard. One shouts: "You're it!" laughing. Sound: splashing water, laughter, a dog barking nearby. Style: bright daylight, home-video feel.

Candid, energetic dialogue with overlapping natural sound — run on Veo 3.1 (veo3).

Slow pan across a rainy city street at night, neon signs reflecting in puddles. No dialogue. Sound: steady rain, a distant car horn, muffled traffic. Style: cinematic, blue and pink neon tones, anamorphic lens flare.

Pure mood and ambience, no speech — run on Veo 3.1 Fast (veo3-fast).

Troubleshooting

Mouths move but the words are wrong or muffled. Shorten the line. One clear sentence lip-syncs far more reliably than two sentences crammed into 8 seconds.

Sound effects are missing. Vague mood words don't render as sound. Replace "tense atmosphere" with the actual objects making noise — a ticking clock, a creaking floor, distant traffic.

The clip is close but not quite right. Don't start over. Reply in the same chat ("make it nighttime," "add rain") — Plurel edits the clip you just made instead of generating from scratch.

You're not sure which tier to use. Use Veo 3.1 Fast for ambience tests and drafts at 20 credits, and save full Veo 3.1 at 85 credits for the take with dialogue and camera moves you actually want to keep — our Veo 3.1 explained guide breaks down when the extra credits are worth it.

For broader camera and motion phrasing that applies beyond Veo, see AI video prompt tips, and if you're comparing models before committing credits, Best AI Video Generator in 2026 lays out the field.

Write the dialogue like a script line, name the sound like a prop list, and Veo 3.1 does the rest.