Veo 3.1

Veo 3.1 has no fixed prompt formula. Pick the elements you need, describe the scene in as much precise detail as you can, and avoid describing again what a connected image already sets.

Prompt structure

Pick only the elements you need from the list below and write in as much detail as you can, without vague wording.

Veo 3.1 has no fixed formula, so you do not need to fill in every element below. Pick only the elements your scene needs.

Subject + Action + Scene and context + Camera angle + Camera movement + Lens and optical effects + Style + Temporal elements + Audio
  • It is best to describe the subject, action, and scene. The other elements are optional and set the mood of the result.
  • Describe the subject specifically rather than with a general noun. Write "a weathered lighthouse keeper in a wool coat" rather than "a person".
  • It is best to write the whole prompt, not just the subject, precisely and without ambiguity.
A weathered lighthouse keeper in a wool coat climbs a spiral iron staircase, one hand on the rail, breath visible in the cold. Interior of a 19th-century lighthouse tower before dawn. Low-angle tracking shot following him upward, wide lens, deep shadows. Cold blue moonlight from above, warm lamp glow from below. Audio: iron steps ringing, wind pressing against the tower, distant surf. No music.

To create a video with Veo 3.1 in the video node, connect a start frame or reference images together with the prompt. Do not describe again in the prompt what a connected image already sets.

You can check the supported inputs, length, ratio, and resolution in Video models.

Dialogue and sound

Veo 3.1 generates sound along with the video, so write dialogue and sound into the prompt separately by type.

Do not put dialogue in quotation marks.

Text in quotation marks can appear on screen like subtitles. For dialogue, write the speaker, add a colon (:), and continue with the line.

A woman says: My name is Clara.

Write each type of sound in its own sentence. It is best to separate sound into three types: sound effects, ambient sound, and dialogue.

TypeExampleWhat you hear
Sound effectthe sound of a phone ringingA phone ringing
Ambient soundthe sounds of city traffic and distant sirensTraffic noise in the city and sirens in the distance
Dialoguethe man in the red hat says: Where is the rabbit?The man in the red hat asking where the rabbit is

Start frame

The start frame already sets the people, background, lighting, and style, so the prompt only needs to say what moves.

Describing again in the prompt what the start frame already shows can lower the quality of the result. Write only what moves and how: the camera, a person, or the background.

What movesExampleWhat happens
CameraSlow dolly in on the subject.The camera slowly moves in toward the subject.
PersonThe character's hair flutters gently in the wind.The person's hair flutters softly in the wind.
BackgroundFog rolls in slowly across the landscape.Fog slowly drifts in across the landscape.
  • Of the three, describing camera movement gives the most stable results.
  • Refer to people as the subject, the woman, he, or she rather than by name.
  • The same applies when you also connect an end frame. Do not describe again what the two frames show; write only the movement between them.
  • It is best to use a sharp, well-shot image as the start frame.

Reference images

Veo 3.1 Pro and Fast use reference images to keep the appearance of people, characters, and products, and the prompt takes no tags.

Veo 3.1 Pro and Fast accept up to 3 reference images. Every connected reference image is used to keep the appearance of a person, character, or product. Reference images cannot be connected together with a start or end frame.

  • Do not write tags or image numbers in the prompt. An image's role is set by which input of the video node it is connected to.
  • Reference images keep only the appearance. Center the prompt on the subject's action and the camera, and it is best to also describe the scene, lighting, and style you want.