How to Write Veo 4 Prompts

Good video prompts work less like slogans and more like compact production briefs. The model is trying to infer direction, camera language, art direction, pacing, and sometimes sound from a few lines of text. Once you start writing with that in mind, prompt quality usually improves very quickly.

Google DeepMind's public Veo guidance still points in a simple direction: the more concrete your visual and cinematic details are, the more controllable the result becomes. Camera framing, motion, lighting, style, environment, subject identity, and audio cues all help reduce randomness. Even while Veo 4 details remain unconfirmed, this is still the safest way to prepare.

A useful six-part prompt structure

• 1. Format: start with the type of video you want, such as a cinematic short, a product ad, a social clip, or a documentary-style scene.
• 2. Subject: define who or what is on screen with stable identity details.
• 3. Action: explain what is happening and how fast or restrained the movement should feel.
• 4. Scene and light: describe the location, time of day, atmosphere, weather, and lighting quality.
• 5. Camera language: specify the shot size, movement, lens feel, and point of view.
• 6. Sound and emotion: add dialogue, ambient sound, effects, and the emotional tone you want the viewer to feel.

Weak prompt vs strong prompt

A weak prompt often sounds like this: a woman walking down the street in a cinematic style. It expresses a genre, but it does not really direct the scene.

Strong prompt example: Realistic cinematic short. At dusk in a narrow Tokyo side street, a woman in her twenties wearing a charcoal trench coat walks slowly across wet pavement while holding a compact camera. The shot begins in medium framing, then gently pushes in toward a half-close-up. Neon reflections break across the ground, light street mist hangs in the air, and distant train noise mixes with footsteps on shallow water. The mood feels restrained and solitary, but not bleak.

The difference is not just length. The stronger version tells the model what must be present and what is supporting atmosphere. It is not more decorative. It is more precise.

How to avoid writing prompts like a list of random details

• Lead with intent, then add detail. Say premium skincare ad first, then describe the bottle, surface, and light.
• Use one main camera movement per prompt unless you are deliberately building a transition.
• Replace references like make it feel like a movie with concrete traits such as low saturation, shallow depth of field, or soft backlight.
• Prioritize stable identity features when describing people: hair, face shape, silhouette, signature outfit colors, and accessories.
• Treat sound as part of the scene, not an afterthought. Audio cues can change the tone of the whole result.

A better way to write product-video prompts

If you are creating product videos, do not begin with brand philosophy. Treat the product like a physical object being photographed. What is the material? How reflective is the surface? Is it sitting on marble, paper, glass, or brushed metal? Generative video models react strongly to material and lighting descriptions, so this part matters.

Product prompt example: Luxury fragrance commercial. A black glass perfume bottle rests on a dark stone surface marked with faint water traces. A cool spotlight falls from the upper left, catching a sharp highlight along the bottle shoulder. The camera descends slowly from above the cap and settles on the front label. The background stays almost black with only a dim blue-gray layer of depth. The soundtrack is a clean glass tap followed by subtle air movement.

Four common mistakes

• Overloading one prompt with too many styles at once.
• Using abstract action words like natural or epic without showing what that actually means on screen.
• Changing character details mid-prompt, which leads to instability.
• Ignoring sound entirely, even though audio cues can shape realism and mood.

A minimal template you can reuse

[Format or style] + [subject identity] + [action] + [scene and lighting] + [camera framing and movement] + [sound or dialogue] + [overall mood]

If you are unsure where to start, use that template for version one, then revise only one variable at a time. That will help you learn whether your prompt problem is really the subject, the camera, the setting, or the emotional tone.