# Veo 3.1 prompt generator

URL: https://dreambench.ai/tools/veo-prompt-generator

A preset of the [AI video prompt generator](https://dreambench.ai/tools/ai-video-prompt-generator): the same engine, set up for this one job.

Veo 3.1 makes clips of up to 8 seconds, widescreen or vertical, and generates sound with every one. This guide shows how a Veo prompt is built and where it goes wrong; its generator writes one: the frame first, a spoken line in quotes after its speaker, what is heard as well as seen, as plain text or JSON.

## What is a Veo 3.1 prompt generator?

It writes a prompt for Google's Veo 3.1 from your choices: one shot of up to 8 seconds, frame first, any line in quotes, the sound, light and camera, in plain text or JSON, asking only for what Veo can make. It is part of every DreamBench plan, from $9.99 a month.

## How to write a Veo 3.1 prompt

1. Describe the shot: Who or what is in frame, what they do and where, and the one line they say, if any. One shot per clip, so keep to one moment.
2. Set the frame and the sound: 4, 6 or 8 seconds, 16:9 or 9:16, then the light, the lens and one camera move. Sound is always on with Veo, so choose ambience alone or with a spoken line.
3. Copy or run it: The prompt opens with the frame and says what is heard straight after the action, as plain text or JSON. The first time, create a free account and choose a plan, then copy it into any tool that runs Veo 3.1, or run it on DreamBench, which offers Veo 3.1 and its Fast and Lite tiers.

## How a Veo 3.1 prompt is built

1. The frame and the length: Widescreen 16:9 or vertical 9:16, and one shot of 4, 6 or 8 seconds, said first, as Google's own examples open.
2. The scene, with its line: Who is in frame, what they do and where, in one sentence. A spoken line goes in quotes straight after the person who says it, so the voice belongs to someone on screen.
3. What is heard: Veo makes sound with every clip, so the prompt says what should be heard: the ambience, the line, each landing in step with the action. Leave it out and Veo makes a soundtrack you didn't choose.
4. The light: The source and its side, and what falls to shadow, so Veo doesn't light the scene flat from the front.
5. The camera in film terms: One shot size, a lens in millimeters and one move: Veo 3.1 understands cinematic language.
6. The look, and nothing written: One register, such as photoreal film, then a closing line against text, captions and watermarks in the frame.
7. The same prompt as JSON: The JSON lays the same words out under names, shot, scene, dialogue, sound, light, camera, look and avoid, so a part is easy to find. Google publishes no JSON format for Veo: it reads JSON as text.

## Example prompts

**A line to camera (Veo 3.1).** An 8-second widescreen shot with one line, in quotes after the man who says it, and the sound around it. Shown as JSON too.

```text
Widescreen 16:9, one 8-second shot. A lighthouse keeper in a thick wool jumper climbs the last steps of a spiral stair and stops at the window, at the top of a lighthouse on a stormy night, and says, "She's coming in fast tonight." Sound: the line clear over the scene's own ambience, every sound landing in step with the action; no music. Light: candle and firelight, warm and unsteady from camera-left, so the subject's camera-right side falls to shadow; the exposure holds the subject; no fill from the camera side. Medium shot on a 50mm lens at a couple of metres: natural proportions, the background softening behind the subject. One move only: a slow push-in that ends close on the subject. Photoreal, shot on film: natural colour, gentle grain, true skin. No text, no captions, no watermark.
```

As JSON:

```json
{
  "shot": "Widescreen 16:9, one 8-second shot",
  "scene": "A lighthouse keeper in a thick wool jumper climbs the last steps of a spiral stair and stops at the window, at the top of a lighthouse on a stormy night",
  "dialogue": "She's coming in fast tonight.",
  "sound": "the line clear over the scene's own ambience, every sound landing in step with the action; no music",
  "light": "candle and firelight, warm and unsteady from camera-left, so the subject's camera-right side falls to shadow; the exposure holds the subject; no fill from the camera side",
  "camera": {
    "framing": "Medium shot on a 50mm lens at a couple of metres: natural proportions, the background softening behind the subject",
    "move": "One move only: a slow push-in that ends close on the subject"
  },
  "look": "Photoreal, shot on film: natural colour, gentle grain, true skin",
  "avoid": "text, captions, watermark"
}
```

**A vertical product moment (Veo 3.1).** A 6-second vertical clip for a phone: no dialogue, so the sound is the scene's own.

```text
Vertical 9:16, one 6-second shot. A barista pours steamed milk into a flat white, drawing a leaf in the foam, at the counter of a small café in the morning. Sound: the scene's own ambience, every sound landing in step with the action; no music. Light: one window's daylight from camera-right, so the subject's camera-left side falls to shadow; the exposure holds the subject; no fill from the camera side. Close-up on an 85mm lens from six metres: the background compressed and soft, only the subject sharp. The camera is locked off; nothing moves but the scene. Photoreal, shot on film: natural colour, gentle grain, true skin. No text, no captions, no watermark.
```

**A wide establishing shot (Veo 3.1).** A wide shot that sets a place: one lateral track, the light from behind, the sound left to the scene.

```text
Widescreen 16:9, one 8-second shot. A night train crosses a snowy mountain pass, its windows glowing, between dark pine slopes under a clear sky. Sound: the scene's own ambience, every sound landing in step with the action; no music. Light: night, lit only by the practical lights in the scene from behind the subject, the camera shooting into the light from the shadow side, a rim of light on the edges; the exposure holds the subject; no fill from the camera side. Wide shot on a 24mm lens from close: the space readable to the edges, the nearest subject looming, lines straight. One move only: a lateral track that keeps pace with the subject, passing a foreground shape. Photoreal, shot on film: natural colour, gentle grain, true skin. No text, no captions, no watermark.
```

## Common mistakes

- Asking Veo for silence: Veo 3.1 makes sound with every clip, so a prompt for a silent one isn't honored. Describe the sound you want instead, and mute the clip in your editor if you must.
- A line with no speaker: A line left at the end of the prompt can go to anyone, or to a voice-over. Put it in quotes straight after its speaker.
- More action than 8 seconds hold: A walk, a talk and a reveal won't fit in one clip. Keep to one moment; a longer scene is several clips, which the composer joins.
- Cuts squeezed into 8 seconds: Google's guide shows timed shots inside one clip, [00:00-00:02] and so on, which leaves each about two seconds. This generator writes one shot; keep to one unless the cut matters.
- Trusting the first take of a short line: Google says natural, consistent speech, especially in short lines, is still being improved. Listen before you use the clip, and run it again if the voice is off.

## Works with

- What it takes: Your own words: who is in the shot, what they do and where. No photo is needed.
- What you get: A prompt as plain text or as JSON.
- DreamBench's composer: Paste it into Video on Veo 3.1, Fast or Lite, which sends your words to the model as written.
- Google Flow and the Gemini app: Both make Veo 3.1 clips from a text prompt.
- The Gemini API and Vertex AI: Send the plain prompt, or the JSON as its text.

## Which model it writes for, and its limits

The Veo 3.1 prompt generator writes for [Veo 3.1](https://dreambench.ai/models/veo-3.1), within what it can make on DreamBench.

| | Veo 3.1 |
|---|---|
| Lengths | 4, 6 or 8 seconds |
| Frames | 16:9 or 9:16 |
| Sound | Always on: ambience, or ambience and a spoken line |
| Shots | One shot, no cuts |
| Dialogue | One line, said in the scene |
| Prompt format | Plain text or JSON |

- Sound written in: Veo 3.1 generates dialogue, effects and ambience with every clip, so the prompt always says what is heard and never asks for a silence the model won't give.
- Its lengths and frames only: The choices stop where Veo does: 4, 6 or 8 seconds, 16:9 or 9:16. A prompt never asks for what the model can't make.
- A line that belongs to someone: A spoken line goes in quotes right after the subject, so the voice belongs to someone on screen.
- Plain text or JSON: The same prompt as a paragraph or laid out under names, whichever is easier to read and change.

**Where it fails.**
- Google says natural, consistent speech, especially in short lines, is still being improved.
- One clip runs 8 seconds at most; a longer scene is several clips.
- Only widescreen 16:9 and vertical 9:16: no square or cinematic frame.
- Google fully supports English prompts; other languages may work but haven't been evaluated.

## Veo 3.1 prompt generator FAQ

- **How do I write a prompt for Veo 3.1?** To write a prompt for Veo 3 or 3.1, open with the frame, widescreen or vertical, and the length, 4, 6 or 8 seconds, then who is in shot, what they do and where, any line in quotes straight after the speaker, what is heard, the light and one camera move. Veo always makes sound, so say what it should be; the generator writes it in that order.
- **What is the best prompt structure for Veo 3?** Google's own guide orders a Veo prompt as cinematography, subject, action, context, and style and ambiance, then directs the sound: dialogue in quotes, effects after SFX: and the background after Ambient noise:. This generator says the frame and the length first, as Google's examples open, then the same parts.
- **How do I make a character speak in Veo 3?** Put the line in quotes straight after the person who says it, as Google's guide does: A woman says, “We have to leave now.” Keep to one short line in 8 seconds, and listen first: Google says short lines are still being improved. For two characters talking, Google builds a conversation as separate shots, one line each; here, make one clip per line and let the composer join them.
- **Does Veo 3 take a negative prompt?** Veo 3 has no separate negative prompt. Google's guide says to describe what you want to exclude positively, a desolate landscape with no buildings or roads rather than no man-made structures. This generator does the same and closes every prompt with one short line: no text, no captions, no watermark.
- **How long should a Veo 3 prompt be?** Google's guide sets no limit and its examples run one to three sentences; 60 to 150 words covers the frame, the scene, the sound, the light and the camera for an 8-second clip, which is what this generator writes. The same prompt runs on Veo 3.1, Veo 3.1 Fast and Veo 3.1 Lite.
- **Can Gemini write a Veo prompt for me?** Yes, and without a plan here. Google's Veo guide suggests asking Gemini to enrich a simple prompt with more descriptive, cinematic language. Gemini doesn't know the lengths and frames DreamBench runs Veo at unless you say; this generator, which needs a DreamBench plan, asks only for 4, 6 or 8 seconds, in 16:9 or 9:16.
- **Should I use a JSON prompt for Veo 3.1?** Use it if it helps you. Google publishes no JSON format for Veo: the model reads JSON as text, like any prompt. Laying the parts out under names makes a long prompt easier to scan, and the generator's JSON carries the same words as its plain prompt, so neither is stronger.
- **Does a Veo 3 prompt work on Veo 3.1?** Yes: the structure is the same. Google says Veo 3.1 builds on Veo 3 with richer sound, truer textures and closer attention to the prompt, so a Veo 3 prompt works, and a precise one gains most. This generator writes for Veo 3.1's lengths and frames as DreamBench runs it.
- **Can I use this for Veo image to video or a first and last frame?** This generator writes a text-to-video prompt. When a picture is the first frame, describe only what moves and what the camera does; the image already sets the subject, the light and the look. On DreamBench, Film Studio gives a Veo 3.1 scene up to three reference pictures, which Google calls Ingredients to Video, to hold a character across clips.
- **Can Veo 3.1 make a clip without sound?** No. Veo 3.1 generates sound with every clip, so the generator has no silent option: choose ambience on its own, or ambience with one spoken line, and describe the sounds you want. For a silent clip, remove the audio in your editor afterward.
- **How long can a Veo 3.1 clip be, and which frames and resolutions does it make?** Up to 8 seconds, in two frames, 16:9 and 9:16, on all three tiers; the generator offers 4, 6 or 8 and opens with the frame; a longer scene is several clips, which the composer joins. The prompt doesn't set the resolution: on DreamBench, Veo 3.1 renders at 720p, 1080p or 4K, Fast and Lite at 720p or 1080p. For a square or 21:9 clip, write for Seedance or Kling.
- **Should I write a Veo prompt in English?** English works best: Google fully supports English prompts for Veo 3.1 and says other languages may work but haven't been evaluated. The generator writes its own sentences in English and passes on your words as you type them, so type the subject, the action and the setting in English. A spoken line in another language can go in quotes, but listen to it first.
- **How much does the Veo 3.1 prompt generator cost?** Each prompt uses 11 credits from your DreamBench plan, and the same prompt again uses none. Starter, $9.99 a month, includes 1,100 credits a month, about 100 prompts, plus AI video and the image tools.
- **Do I need an account to use the Veo 3.1 prompt generator?** Yes: a free account and a DreamBench plan. The tools are part of every plan, from Starter at $9.99 a month.

Calls no model; 11 credits a prompt. DreamBench is an independent platform; its tools come with every plan (https://dreambench.ai/pricing). Copy last checked 2026-10-14.