Veo prompt guide: video with sound and dialogue

What Veo 3.1 can do

According to the Gemini API documentation, Veo 3.1 makes 4, 6 or 8 second clips in 16:9 or 9:16 at 24 frames per second. 720p works at any length, while 1080p and 4K are limited to 8 second clips. Audio is always on.

The model accepts up to three reference images for a character or object, a start frame and a last frame. Veo 3.1 Lite drops references and 4K, which makes it a good fit for cheaper drafts.

Prompt structure

Google recommends building the prompt from components. Subject and action are required, the rest adds precision. Put audio in separate sentences after the visuals.

• Subject: be specific, "an elderly craftsman in a leather apron" rather than "a man"
• Action: what happens and how, with emotion and pace
• Scene: location, time of day, weather, set details
• Camera: shot size, angle and movement
• Style: lighting, mood, art style, color
• Audio: dialogue, sound effects, ambience and music

Writing dialogue

Put each line in quotes and always say who speaks and in what tone. An 8 second clip fits one or two short phrases. A long monologue gets cut off or rushed.

Google lists English as fully supported, other languages work with varying quality. Test lines in other languages on a short clip, and add a voice over in editing when the wording has to be exact.

• "Medium shot. A barista sets a cup on the counter, smiles and says warmly: “Your coffee is ready”. Low cafe chatter, clinking dishes, no music"
• "Close up. A young woman looks into the camera and whispers: “It is a secret”. Rain taps on the window"
• "Two friends in a kitchen. The first asks, surprised: “Did you cook this yourself?” The second nods proudly: “Of course”"

Sound effects and ambience

Describe sounds directly and separately from the image: footsteps on gravel, a creaking door, surf on the shore. Ambience sets the mood of the scene, such as city hum or birdsong. If you do not want music, say so.

If the model form has a separate exclusion field, Google advises listing the unwanted elements themselves without words like "no" or "do not". For example, write "urban background" there instead of "no urban background".

Camera and style

Veo understands standard cinematography terms. Tell a dolly, where the camera moves toward the subject, apart from a zoom, where the lens changes. Time can be directed in words too: slow motion, time lapse, a gradual change.

• Shots: wide, medium, close up, extreme close up, over the shoulder, point of view
• Angles: eye level, low angle, high angle, bird’s eye view
• Moves: static, pan, tilt, dolly, crane, drone, handheld, arc around the subject
• Optics: shallow depth of field, rack focus, wide angle, telephoto

From an image, between frames and with references

In image to video the photo becomes the first frame, so pick the image closest to how the scene should open. Then describe what happens next.

A last frame sets the ending, and Veo builds the transition between the two images. References keep a character or product consistent when you describe the shot itself in text.

How to do it in Solistra

The Solistra catalog has Veo 3.1, Veo 3.1 Fast and Veo 3.1 Lite. The price depends on the version, duration and resolution and is shown before you run it.

• Open Video and choose Veo 3.1 or one of its faster versions
• Upload a start frame or references if you need them
• Write the prompt with separate sentences for dialogue and sound
• Pick duration and resolution, check the price on the button and run it