A good Veo 3.1 prompt reads like a short shot description: who is in the frame, what they do, where they are, how the camera behaves, what it should look like and what we should hear. Veo 3.1 follows prompts more closely than Veo 3 did, according to Google, so the more precisely you describe the shot, the more of it you get back.
This guide covers how to structure a prompt for one 8-second clip, which camera and style words are worth using, how to write dialogue and sound, how start and end frames change what the prompt needs to say, and ten prompts you can paste and adapt.
Before you write: what one Veo 3.1 clip can hold
On Veo Video, every Veo 3.1 generation is an 8-second clip with sound, in 16:9 or 9:16, at 720p, 1080p or 4K. You can start from text, from a photo (it becomes the first frame) or from a photo plus an end frame. Reference images run on Veo 3.1 Fast, with up to three per clip.
Eight seconds is roughly one shot with one clear action, or two or three short beats. It is enough for a spoken line or two, not a paragraph of dialogue. Most weak prompts try to fit a whole scene into one clip; the fix is usually to cut the prompt down to the single moment you want.
The structure of a Veo 3.1 prompt
Google's Veo prompt guidance lists the elements that shape a clip: subject, action, style, camera positioning and motion, composition, focus and lens effects, and ambiance. Veo 3.1 also generates sound, so audio belongs in the prompt too. A practical order:
| Element | What it controls | Example |
|---|---|---|
| Shot and camera | Framing and movement | "Close-up, slow dolly in" |
| Subject | Who or what is in frame | "A woman in her 30s in a soft grey sweater" |
| Action | What happens during the 8 seconds | "takes her first sip of coffee and smiles" |
| Scene | Place, time, weather | "on a small balcony above a quiet street at dawn" |
| Style and light | Look and mood | "warm morning light, shallow depth of field" |
| Sound | Dialogue, effects, ambience | "birds chirping softly, a distant city hum" |
Put together:
Close-up, slow dolly in on a woman in her 30s in a soft grey sweater, taking her first sip of coffee on a small balcony above a quiet street at dawn. Morning light grazes her face. Birds chirp softly, a distant city hum. Warm color grade, shallow depth of field, TV commercial style.
You do not need every element in every prompt. Anything you leave out, the model fills in, which is fine for details you do not care about and a problem for details you do.
Camera: shot types and movement
Camera language is one of the most reliable levers in a Veo prompt. Name the framing, then name one movement. Eight seconds rarely has room for more, so keep to one move rather than stacking several ("pan left while craning up and zooming in").
| Shot | Use it for | Words to use |
|---|---|---|
| Wide / establishing | Places, scale, context | "wide shot", "establishing shot" |
| Medium | People talking, general action | "medium shot", "waist-up" |
| Close-up | Faces, emotion, product details | "close-up" |
| Extreme close-up | Texture, small objects | "extreme close-up", "macro shot" |
| POV | First-person immersion | "POV shot", "first-person view" |
Movements that work well in a single clip:
- Dolly in / dolly out: the camera moves toward or away from the subject; good for building tension or revealing context.
- Pan: the camera turns horizontally; good for scanning a scene or following sideways motion.
- Tracking shot: the camera travels alongside the subject.
- Crane shot: the camera rises or descends; good for reveals.
- Orbit: the camera circles the subject.
- Static: a locked-off camera, useful for dialogue and product shots.
Lens and focus words help too: "shallow depth of field", "rack focus from the cup to her face", "wide-angle lens", "macro lens", "shot on 35mm film" for grain and warmth.
Dialogue and sound
Veo 3.1 generates sound with every clip. If you say nothing about audio, it still builds a soundtrack for the scene; if you describe it, you get more control. Google's guidance is to put speech in quotes, describe sound effects explicitly and describe the ambience separately.
Dialogue
Put the spoken line in quotes, say who speaks and how:
Medium shot of a tired detective behind a desk in a dim office, rain on the window behind him. He looks up and says in a weary voice, "Of all the offices in this town, you had to walk into mine." Film noir look, hard shadows from window blinds.
- Keep it to one or two short sentences per clip. Anything longer will not fit comfortably in 8 seconds.
- With two speakers, name them and give each a line, for example Host A: "Welcome back!" Host B: "Glad to be here."
- Describe the voice ("weary", "excited whisper", "calm and slow") so the delivery matches the picture.
Sound effects and ambience
Tie each sound to something visible, and write the effects the way you want them heard:
Wide shot of a narrow alley under buzzing neon signs as cold drizzle falls. A distant alarm wails, neon hums, rain patters on metal bins.
The banana-peel prompt also shows how to handle a sequence: it puts the actions in order and marks the turns with words like "suddenly" and "in the next moment". That gives the model a timeline to follow inside the 8 seconds.
Style, color and light
Style words work best when they describe something visible rather than a vague quality.
| Look | Words that help |
|---|---|
| Cinematic | "shot on 35mm film", "anamorphic lens", "film grain" |
| Documentary | "handheld camera", "natural light", "documentary style" |
| Horror | "desaturated colors", "heavy grain", "low angle", "flickering light" |
| Sci-fi | "neon-lit", "holographic display", "cool blue light" |
| Commercial | "TV commercial style", "clean studio lighting" |
| Animation | "hand-drawn anime style", "cel-shaded", "vibrant colors" |
For light and color, name the direction and the time of day: "dramatic side lighting", "backlit silhouette", "golden hour", "blue hour", "overcast diffused light", "warm golden tones", "cyan-magenta grade".
Describe what you want rather than listing what you don't. "A clean, unlabeled glass bottle on a plain black surface" gets you further than a list of things to avoid.
Starting from images
Image to video
When you upload a photo, it becomes the first frame. The prompt no longer needs to describe what the subject looks like; spend it on the motion, the camera and the sound instead. For example: "She turns toward the window and smiles, slow dolly in, soft room tone and a distant car passing."
Start and end frames
Add an end frame and Veo 3.1 builds the motion between the two images. Use it when a shot has to land on a known composition: a reveal, a product turning to face the camera, a transformation. Keep the prompt focused on how the change happens ("the paper ball unfolds and folds itself into a paper airplane"), since both endpoints are already fixed.
Reference images (Veo 3.1 Fast)
Reference images are different from a first frame: they tell the model what a person, character or product looks like, and the model places them in a new scene. On Veo Video you can add up to three on Veo 3.1 Fast. The prompt then describes the scene and action; the images carry the identity.
10 Veo 3.1 prompts you can paste and adapt
Each prompt is written for one 8-second clip. The suggested aspect ratio is a starting point; all of them can run in either 16:9 or 9:16.
1. Product reveal (16:9)
Close shot of a sleek black smartwatch resting on a rugged rock at a mountain cliff edge. The camera pulls back and rises in one smooth drone move, revealing a vast alpine valley at sunrise. Wind rushes softly, a low cinematic swell. Product commercial style, crisp natural light.
2. Speaker on stage (16:9)
Medium shot of a confident speaker at a podium in a modern conference hall, static camera. She gestures naturally and says, "The future of AI is not replacement. It is collaboration." Soft stage lighting, light audience murmur, polished corporate look.
3. Nature documentary (16:9)
Wide tracking shot following a lone wolf trotting through fresh snow in a dense pine forest at dusk. Snow crunches under its paws, wind whispers through the trees. Documentary style, natural light, 35mm film grain.
4. Horror hallway (9:16)
Low-angle wide shot down a long, empty hospital hallway with flickering fluorescent lights. A lone figure at the far end walks slowly toward the camera, footsteps echoing. A light buzzes and cuts out. Desaturated colors, heavy grain.
5. Coastal ride POV (9:16)
POV shot from a motorcycle helmet on a winding coastal road, leaning into a curve with cliffs and a sparkling ocean below. Engine roar, wind rushing past. Golden hour light with sun flares, high-energy action sports style.
6. Chef at work (16:9)
Close-up of a chef's hands arranging fresh herbs on a marble counter, then the camera tilts up to her focused face. Knife taps, a pan sizzling in the background. Soft overhead window light, warm lifestyle look.
7. Fashion walk (9:16)
Slow dolly in on a model walking through an empty white art gallery in a flowing silk dress; each step sends ripples through the fabric. Heels click on polished concrete, quiet gallery echo. Soft diffused light, high-fashion editorial style.
8. Quiet emotional moment (16:9)
Medium shot of an elderly man on a park bench feeding pigeons in warm afternoon light under autumn trees. He pauses, looks up and smiles as leaves drift past. Pigeons coo, leaves rustle, distant children's laughter. Nostalgic tone, shallow depth of field.
9. Sci-fi interface (16:9)
Close-up of hands swiping and pinching a transparent holographic display, rotating a glowing 3D globe of data. Blue-white light from the interface illuminates the face behind it. Soft electronic chimes with each gesture. Clean, futuristic look.
10. Luxury product (9:16)
Extreme close-up of a luxury perfume bottle on a reflective black surface under a single dramatic spotlight. The bottle slowly rotates, catching golden highlights on its facets. A soft, resonant chime. Premium commercial style.
A workflow from draft to final
- Start short. Write the shot, the subject and the action in one or two sentences and run it on Veo 3.1 Lite at 720p. It is the cheapest way to check that the model understood the idea.
- Add the camera and the look. Once the action is right, add one camera move, the lighting and the style.
- Add the sound. Write the dialogue in quotes and name the effects and ambience.
- Render the keeper. When the prompt works, run it at 1080p or 4K, on Lite or on the full Veo 3.1 tier for the most detailed take. Each run is a new take, so expect a close match to your draft rather than an identical clip.
Tips along the way:
- Change one thing at a time between runs (the camera or the light, not both) so you can see what made the difference.
- Run a prompt that works a few times; each generation is a different take, and one will usually be better than the rest.
- Write prompts in English for the most predictable results.
- For sequences longer than 8 seconds, write each shot as its own prompt and reuse the same character, palette and style words across them.
Common mistakes
| Mistake | What happens | Fix |
|---|---|---|
| Several camera moves in one prompt | Competing instructions in a short clip | One framing, one movement |
| A whole scene in one clip | Rushed or skipped actions | Keep to one moment or two or three beats |
| Long dialogue | Lines get cut or rushed | One or two short sentences |
| Vague subject | Random details filled in | Say age, clothing, key features |
| No sound direction | A generic soundtrack | Name effects and ambience; quote the dialogue |
| Different style words across related clips | Clips that don't cut together | Reuse the same palette and style phrases |
| Real public figures or photos of children | The request is blocked | Use fictional characters and adults |
If a generation is blocked by Google's safety filters, the credits come back to your balance.
Try it
New accounts on Veo Video get 30 free credits, enough for one 8-second Veo 3.1 Lite clip at 720p. Pick one of the prompts above, adjust it to your subject and see how the model reads it.
For more on the model itself, read Veo 3 vs Veo 3.1: what changed.