How should you structure a Veo 3.1 prompt?

Subject and action first, then camera movement, then environment and lighting, then audio. Veo 3.1 follows directorial language well — write like a shot description in a screenplay, not like an image prompt full of style adjectives.

Can Veo 3.1 generate dialogue?

Yes — Veo 3.1 generates synced speech. Put the exact line in quotes in your prompt ('she says: "We need to leave now"') and describe the delivery (whispered, excited, deadpan). Quoted lines produce noticeably better lip sync than paraphrased descriptions.

Does Veo 3.1 support image-to-video?

Yes, and it is the most reliable workflow: start from a still image (a photo or a generated frame) and prompt only the motion and audio. Identity, styling, and composition come from the image; the prompt directs what happens next.

Why do my Veo 3.1 results look generic?

Usually the prompt describes a mood instead of an event. Replace adjectives with actions and camera direction: instead of 'cinematic emotional scene', write 'slow push-in on her face as she reads the letter, her expression shifts from confusion to relief, soft window light, quiet room tone.'

Back to blog

Veo 3.1 Prompt Guide: How to Write Prompts That Work [2026]

July 3, 2026By Bilal Azhar

How to prompt Google's Veo 3.1 for video — structure, camera language, audio and dialogue cues, image-to-video, and copy-paste example prompts that actually work.

Veo 3.1 prompts work best written like director's notes: one clear action, explicit camera movement, concrete lighting, and quoted dialogue for synced speech. Use Veo 3.1 on Morphed with its Fast variant for drafts, and image-to-video for identity control.

Veo 3.1 is the most literal-minded of the frontier video models. That's a compliment: it does what you say, including camera moves, timed actions, and spoken lines. It also means vague prompts produce vague results — the model won't invent a great shot from "cinematic, emotional, 8k."

This guide covers the prompt structure that reliably works, Veo-specific strengths (dialogue, audio, physics), and copy-paste examples by use case.

Key Takeaways

Write events, not moods — one clear action per clip
Use explicit camera language: push-in, dolly, orbit, handheld, static
Put dialogue in quotes for synced speech
Describe sound — Veo 3.1 generates audio, and prompted audio beats default ambience
Use image-to-video when identity or product accuracy matters

The Prompt Structure

The reliable order:

[Shot type + camera movement] + [subject + one action] + [environment + lighting] + [audio] + [style notes if needed]

Example:

Medium close-up, slow push-in. A man in his 50s at a workbench looks up
from a pocket watch he's repairing and smiles slightly. Warm tungsten
desk lamp, dust in the light beam. Audio: ticking clocks, a chair creak,
soft workshop ambience.

Every element earns its place: the camera has a job, the subject has one action with a beginning and end, the light source is named, and the audio is concrete.

Camera Language Veo Understands

Veo 3.1's instruction-following on camera direction is the best reason to use it. These all work reliably:

Instruction	Result
`static shot` / `locked-off camera`	No drift — important for product shots
`slow push-in` / `slow dolly forward`	Cinematic emphasis move
`tracking shot beside her`	Lateral follow
`orbit around the subject`	Arc move
`handheld, slight shake`	Documentary/UGC feel
`crane up and reveal`	Rising establishing shot
`rack focus from foreground to background`	Focus pull

One camera instruction per clip. Stacked moves ("push in while orbiting and craning") degrade into mush.

Dialogue and Audio

Veo 3.1 generates synced speech and ambient sound — its biggest practical differentiator.

Dialogue: quote the exact line and direct the delivery.

A barista slides a cup across the counter, grins, and says: "Third one
today — rough morning?" Casual, teasing tone. Coffee shop ambience,
milk steamer hissing in the background.

Sound design: name specific sounds instead of "with audio." Footsteps on gravel, rain on a windshield, a distant train — concrete Foley cues anchor the clip's timing and realism.

Music: prompting music styles works but is hit-or-miss; most creators mute generated music and add licensed tracks in the edit.

Image-to-Video: The Control Workflow

When a specific person, product, or composition must stay accurate, start from a still and prompt only motion and audio:

Generate or upload the frame — e.g. a portrait made with Nano Banana Pro, a product shot, or a virtual try-on result.
Prompt what happens: "She turns toward the window and laughs softly. Camera static. Audio: distant street noise, her quiet laugh."
Iterate on motion only — identity is already solved.

This is the backbone of every professional workflow we cover: UGC ads, real estate, e-commerce, and fashion.

Copy-Paste Prompts by Use Case

Talking-head hook (UGC ad):

Casual selfie-style video, handheld phone camera feel. A woman in her
late 20s in a bright kitchen holds up a glass water bottle and says:
"Okay, whoever invented this deserves a raise." Enthusiastic, genuine.
Natural window light. Audio: her voice, quiet kitchen ambience.

Product lifestyle:

Static macro shot. A hand pours cold brew coffee over ice in a clear
glass, condensation on the surface. Morning light from the left, shallow
depth of field. Audio: ice clinking, liquid pouring, soft cafe ambience.

Real estate interior:

Slow, smooth dolly forward from the entryway into a bright living room
toward floor-to-ceiling windows. Steady gimbal motion, natural afternoon
light, no people. Audio: quiet room tone, birds faint outside.

Cinematic narrative:

Wide shot, slow crane up. A woman stands alone on an empty beach at dawn,
her coat moving in the wind, waves rolling in. Cold blue light with a warm
horizon line. Audio: wind, waves, a single gull.

Documentary/interview style:

Medium shot, locked-off camera, shallow depth of field. An elderly
fisherman sits by his boat mending a net, looks up at the interviewer
off-camera and says: "The sea gives you exactly what you respect it for."
Weathered voice. Golden hour harbor light. Audio: rope creak, water lapping.

Common Failure Modes and Fixes

Symptom	Cause	Fix
Generic drifting footage	Mood prompt, no event	Give one action with a start and end
Bad lip sync	Paraphrased dialogue	Quote the exact line
Warping mid-clip	Too much motion requested	One camera move, slower speed
Wrong person/product	Text-to-video for specifics	Switch to image-to-video
Flat sound	No audio prompt	Name 2–3 concrete sounds

Veo 3.1 vs the Alternatives

Veo 3.1's edges are dialogue, audio, physics, and instruction-following. Kling V3 beats it on 4K output and some fabric/product motion; Sora 2 beats it on surreal world-building; Seedance 2.0 beats it on speed and cost. Detailed matchup in Seedance 2.0 vs Veo. On Morphed you can run the same prompt across all of them and compare.

Frequently Asked Questions

What's the ideal Veo 3.1 prompt length?

Long enough to specify one shot completely — usually 2–4 sentences. Longer prompts describing multiple scenes confuse a single generation; break scenes into separate clips.

Can Veo 3.1 do vertical video?

Yes — 9:16 output works well for TikTok and Reels content. Compose for vertical in the prompt ("vertical framing, subject centered").

Fast vs standard — when to use which?

Draft with the Veo 3.1 Fast variant to find the right prompt, then rerun winners on standard for final quality. The prompt transfers directly — both run on Morphed's Veo 3.1 page.

Where can I use Veo 3.1?

Veo 3.1 runs on Morphed with free signup credits, alongside Kling, Sora 2, Seedance, and the rest of the frontier models.

Direct your first Veo 3.1 shot now. Try Morphed free →