Veo 3.1 Prompts with Dialogue and Sound: A Practical Guide

Why sound belongs in the prompt

Veo 3.1 is Google's cinema-grade video model on Biggsfield, built for cinematic realism with natural, synchronized dialogue and sound. You do not add audio afterwards. The voices, room tone and effects come out of the same generation as the picture, timed to what happens on screen.

That changes how you write. A silent-video prompt can stop at the camera and lighting. A Veo prompt that ignores audio leaves the soundtrack to chance. The model will usually fill the gap with something, but not necessarily what you need. If a door should slam on the cut or a character should whisper rather than speak, you have to say so.

Both Veo 3.1 and Veo 3.1 Fast offer 4, 6 and 8 second durations, render at 720p, 1080p or 4K, and generate sound with every clip. Each generation is one continuous shot of up to 8 seconds, so think of every prompt as a single take, not a sequence.

The structure Veo understands

Veo responds best to prompts written in a consistent order: subject and action, then setting, then camera (shot size, lens, movement), then lighting, and finally audio. Keeping audio last makes it easy to read and edit, and it keeps the visual description clean.

Subject and action: who is on screen and what they physically do. Be concrete. "A woman in a rain-soaked trench coat sets her phone face down on the bar" works better than "a mysterious woman at a bar."

Setting: where and when. A late-night diner, a quiet library at closing time, a rooftop at dawn. The setting also tells the model what ambience to expect, so the two should agree.

Camera: shot size, lens and movement. Medium close-up, 50mm, slow push-in. Because Veo works in one continuous shot, choose a single movement that serves the moment rather than stacking several.

Lighting: direction and quality. Warm practical lamps, hard noon sun, cool window light from camera left.

Audio: dialogue, sound effects and ambience, each written in its own clearly marked line. The next two sections cover how to format them.

Writing dialogue in quotes

Veo expects spoken lines in this pattern: the character, the word says, a colon, the line in double quotes, and the delivery in parentheses. For example: The bartender says: "We closed ten minutes ago." (flat, tired, not looking up).

Name the speaker by something visible. If there are two people in frame, "the older man in the grey suit" and "the young woman by the window" are far easier for the model to match to faces than generic labels. This is what keeps lip sync attached to the right person.

The delivery note matters as much as the words. Volume, pace and emotion (whispered, rushed, half-laughing, through gritted teeth) steer how the line sounds and how the face moves while saying it.

Keep lines short. A clip lasts 4 to 8 seconds, and natural speech needs room to breathe. As a rule of thumb, one or two short sentences per clip feels unhurried, and two characters trading a single line each is about as much exchange as fits comfortably in 8 seconds. If your script runs longer, split it into several clips and assemble them later.

Write lines the way people talk. Contractions, fragments and interruptions sound more natural than complete, written-style sentences.

Ambience and sound effects

Beyond dialogue, Veo takes two kinds of audio cue: SFX for specific, timed sounds, and Ambient for the continuous bed underneath the scene. Mark them with the labels SFX: and Ambient: so the model knows which is which.

Use SFX for sounds tied to an action you described: a coffee cup set down on a saucer, keys dropped on a tile floor, a car door shutting. Tie each one to the visual moment so it lands in sync. If the action is not in your visual description, the sound has nothing to attach to.

Use Ambient for the room or world: distant traffic and light rain, a busy restaurant with clattering cutlery, wind across an open field, the hum of fluorescent lights. Ambience is what makes a shot feel located in a real place, and it covers the silence between lines of dialogue.

Fewer, sharper cues beat long lists. One or two sound effects and one ambient description is usually enough for a clip of this length. Piling on ten sounds tends to produce a muddy mix where nothing stands out.

If you want quiet, say so. "Ambient: near silence, only a ticking wall clock" is a valid and often powerful direction for tension.

Veo 3.1 or Veo 3.1 Fast

Veo 3.1 Fast offers Veo quality, faster and cheaper, and it uses exactly the same prompting: subject and action, setting, camera, lighting, then audio with quoted dialogue and SFX and Ambient cues. Anything you write for one works for the other.

The price difference is significant. On Biggsfield, a 4 second clip at 720p or 1080p with sound costs 267 credits (about $2.57 on Creator) on Veo 3.1 and 100 credits (about $0.96 on Creator) on Veo 3.1 Fast. A 6 second 1080p clip is 401 credits (about $3.86) on Veo 3.1 versus 150 credits (about $1.44) on Fast. At 4K, a 4 second clip is 400 credits (about $3.85) on Veo 3.1 and 200 credits (about $1.92) on Fast.

Use Fast for drafting. Testing whether a line reads well, whether the delivery note lands, or whether a camera move suits the moment is iteration work, and Fast lets you run several versions for the price of one full Veo 3.1 render.

Use Fast for volume. Social clips, multiple variations of a hook or a batch of short scenes are a natural fit, especially when speed matters.

Move to Veo 3.1 for the final take when a shot carries the piece: the hero moment of an ad, a key line of dialogue, a close-up where performance and sound quality are under scrutiny. Because the prompt format is identical, switching is just a matter of re-running the version you have already refined.

Biggsfield's AI Video studio also lets you run one idea on up to four models at once and keep the best take, which is a simple way to put Veo 3.1 and Veo 3.1 Fast side by side on the same prompt and judge the difference yourself.

Four example prompts

Each example follows the same order: subject and action, setting, camera, lighting, audio. Adjust durations to fit the amount of dialogue.

8 seconds. A middle-aged waitress in a faded blue uniform refills the coffee of a young man in a wet denim jacket, then pauses and leans on the counter. A small roadside diner at 2 a.m., empty booths, rain streaking the front window. Medium two-shot, 35mm lens, slow push-in toward the counter. Warm overhead pendant lights, cool blue neon spilling through the window. The waitress says: "You've been nursing that cup for an hour, hon." (gentle, amused). The young man says: "Waiting for someone." (quiet, embarrassed). SFX: coffee pouring into a ceramic mug. Ambient: steady rain on glass, low hum of a refrigerator.

6 seconds. A woman in her thirties with a linen apron holds up a fresh sourdough loaf to the camera and taps the crust with one finger. A small bakery kitchen in early morning, flour on the steel worktable, bread racks behind her. Medium close-up, 50mm lens, static camera at eye level. Soft daylight from a window camera left, gentle warm fill. The baker says: "Hear that? That's how you know it's ready." (warm, proud, slightly conspiratorial). SFX: two hollow knocks on the crust, timed to her finger taps. Ambient: quiet oven fan, faint birdsong outside.

8 seconds. A cyclist in a yellow rain jacket rides through a narrow cobblestone street and rings her bell as she passes a café terrace. An old European town at dusk after rain, wet stones reflecting shop lights. Wide tracking shot, 24mm lens, camera moving alongside the cyclist at walking height. Low golden streetlights, deep blue sky, glistening reflections. No dialogue. SFX: bicycle tires hissing over wet cobblestones, a bright double ring of the bell. Ambient: distant church bells, muffled café chatter, water dripping from awnings.

4 seconds. A man in pajamas opens the fridge, freezes, and slowly turns his head toward the camera with wide eyes. A dark apartment kitchen late at night. Close-up, 85mm lens, static, framed from inside the fridge looking out. The only light is the cold white glow of the fridge interior on his face. The man says: "Who ate my cake?" (whispered, deadly serious). SFX: the fridge door seal unsticking as it opens. Ambient: fridge compressor hum, a clock ticking in another room.

Putting longer scenes together

Because every Veo generation is a single shot of up to 8 seconds, a conversation or a short ad is built from several clips. Write each prompt so it stands on its own, repeat the key visual details of characters and setting in every one, and keep the ambience consistent so the cuts feel like the same place.

Once you have your takes, Biggs Cut sits next to your generations as a timeline editor: trim and arrange clips, add music and voiceover on separate tracks, burn in word-timed auto subtitles, and export 9:16, 1:1 or 16:9 from the same project. Rendering happens in the cloud, so you can close the tab while it finishes.

Questions

More guides