GenMagic logoGenMagic
Sign in

Veo 3.1 Prompt Guide: Dialogue, Sound and Camera, With Examples You Can Run (2026)

Live figures last changed · 5 min read · The GenMagic team

Veo 3.1 is the model to use when a clip has to speak: put the exact line in quotes, say who says it and how, describe the sound around it, and keep the shot to 4, 6 or 8 seconds. Everything else follows the usual order for a shot: style, subject, action, setting, camera, light. This guide shows how, with four prompts you can open in the studio and run as they are.

This guide covers the structure, dialogue, sound, camera and shot, opening on your own picture, what to avoid and examples.

A note on who wrote this: GenMagic runs Veo 3.1 Lite, Fast and the full Veo 3.1 and writes from the prompts its users run on them. The three tiers compared on one spoken ad: Veo 3.1 vs Fast vs Lite. Where to get Veo 3.1 without a subscription: the Veo 3.1 guide.

Table of contents

  1. The structure
  2. Dialogue
  3. Sound
  4. Camera and shot
  5. Opening on your own picture
  6. What to avoid
  7. Examples you can run
  8. Questions

The structure

Write one shot, in this order:

  1. Style: "Photoreal", "Handheld vlog", "Cinematic product shot", "Stylized 3D animation".
  2. Subject: who, with the traits that matter (age, clothes, mood).
  3. Action and line: what they do, and the line in quotes.
  4. Setting: where and when.
  5. Camera: shot size and movement.
  6. Light and sound: the light, and what we hear besides the voice.
  7. Length: 4, 6 or 8 seconds, the lengths Veo 3.1 makes.

Dialogue

  • Put the exact words in quotes and say who speaks: The barista says warmly: "Your usual, extra foam."
  • Say how it is said: warmly, out of breath, whispering, laughing.
  • Fit the line to the length: about 2 to 3 words a second; an 8-second clip holds one or two sentences.
  • Two speakers: name each line's speaker in order: The man asks: "..." The woman says: "...".
  • English works most reliably; test another language on a short clip first.

Sound

Veo 3.1 makes the sound with the picture. Describe what we hear besides the voice: "quiet cafe sounds", "wind noise", "distant traffic", "a music box". Without a description, it chooses one; with one, it follows it.

Camera and shot

  • Shot size: close-up for a face and a line, medium shot for a person and their hands, wide shot for a place.
  • One movement: slow push-in, orbit, tracking shot, handheld, or a static shot.
  • Vertical or horizontal: Veo 3.1 makes 16:9 and 9:16; choose 9:16 for Shorts, Reels and TikTok.
  • Resolution: 720p and 1080p on all three tiers, 4K on Fast and the full Veo 3.1.

Opening on your own picture

Add a picture as the first frame and the clip starts on it: a product photo for an ad, a portrait for a talking photo, a frame from your footage for an insert. A last frame sets where the clip ends. Then the prompt only needs the motion, the line and the sound: the picture already gives the look. More: talking photo, image to video.

What to avoid

  • Too many words for the length: a line that does not fit is cut or rushed.
  • Several scenes in one prompt: split them into clips and continue from the last frame.
  • Readable text in the frame: add titles in your editor.
  • Real people's names, brands and existing characters: refused, and not yours to use.
  • AI people presented as real customers: in an ad, an AI person is a presenter, never a customer giving a testimonial.

Examples you can run

Each link opens the GenMagic studio with the model and the prompt ready; you see the price before you generate. Veo 3.1 Lite costs $0.075 a second at 720p on GenMagic today.

A spoken line (Veo 3.1 Lite): A barista in a cozy cafe looks into the camera and says warmly: "Your usual, extra foam." She slides the cup across the counter and smiles. Medium shot, soft window light, quiet cafe sounds, 6 seconds. Open in the studio

Two speakers (Veo 3.1 Lite): Two friends sit on a park bench in autumn. The man asks: "Did you really quit your job?" The woman laughs and says: "Best decision I ever made." Static medium shot, golden hour light, leaves falling, distant traffic, 8 seconds. Open in the studio

A product shot with sound (Veo 3.1 Fast): Cinematic product shot: a matte black coffee dripper on an oak counter, water pours in a slow spiral, steam rises, the camera slowly orbits, soft morning light, shallow depth of field, gentle pouring sound, 8 seconds. Open in the studio

A vlog moment (Veo 3.1 Lite): Handheld vlog shot: a hiker on a windy mountain ridge turns to the camera and says, out of breath: "Okay, we made it. Look at this view." She pans the camera to the valley below. Wind noise, bright daylight, 8 seconds. Open in the studio

For prompts on other models: Seedance, Wan 3.0, Kling 3.0, Hailuo 3.

Teaching AI video or images on YouTube, in a course or in n8n templates? The creator program pays 15% in cash of what the people you bring pay in their first month.

Questions

How do I make Veo 3.1 say a specific line? Put the exact words in quotes and say who speaks them and how: She says warmly: "...". Keep the line short enough for the clip's length.

How long can a Veo 3.1 clip be? 4, 6 or 8 seconds per clip. For longer scenes, make several clips and open each on the last frame of the one before.

Which Veo 3.1 tier should I use? Lite for most clips, Fast for sharper motion and 4K, the full Veo 3.1 for the finest detail. On one spoken ad, Lite matched the full model at an eighth of the price: the test.

Can Veo 3.1 keep the same character across clips? It takes no reference images, so open each clip on a frame of your character (a first frame) and describe the character the same way each time. For reference images, use Seedance 2.0 or Wan 3.0.