How to Write Veo 3 Prompts

Veo 3 and Veo 3.1 reward the same thing: one clear scene, described the way a film crew would describe it. This guide gives you the structure, the vocabulary and the mistakes to avoid.

The 5-part structure

Every strong Veo prompt answers five questions in order:

Subject + action — who or what, doing one thing. One scene, one beat.
Camera — how the shot is captured: static, dolly in, orbit, handheld, crane, FPV.
Lighting — golden hour, neon night, soft studio, candlelit, volumetric rays.
Style — 35mm cinematic, documentary, anime, claymation, vintage film, hyperreal.
Audio — ambient sound, one dialogue line in quotes, voiceover, or music-driven.

Assembled into one sentence, that's a complete prompt — exactly what the free generator builds for you.

Subject: one scene, one beat

The most common failure is stacking three scenes into one prompt ("a chef cooks, then walks outside, then the city timelapses"). Veo generates roughly one shot — give it one moment with a visible beginning and end: "a chef tosses noodles in a flaming wok, sparks flying toward the lens." Concrete nouns and physical verbs beat adjectives.

Camera: name the move

Veo understands real film vocabulary. "Slow dolly in" produces a forward glide; "orbit around the subject" gives you the hero-product arc; "handheld follow" adds documentary energy. If you want the scene to speak for itself, say "locked-off static tripod shot" — a deliberate choice, not a missing one.

Lighting: the cheapest cinematic upgrade

Lighting does more mood work than any filter. Golden hour flatters everything. Neon night + wet streets reads instantly as "city at night". Volumetric rays through haze add depth to any interior. Pick a lighting phrase that matches the emotion you want before you touch style.

Audio: Veo 3's superpower

Veo 3 and 3.1 generate native audio, so describe it: ambient sound ("rain on the awning, distant traffic"), or exactly one short dialogue line — put the words in quotes in your subject. Long monologues degrade lip-sync; one natural line lands best. If you'll add your own sound in the edit, say "no audio".

Five mistakes to avoid

Multiple scenes in one prompt — split them into separate generations.
Negative instructions — "no blur, no distortion" often invokes what you banned. Describe what you do want ("crisp, in focus").
Adjective soup — "amazing epic stunning 8K masterpiece" does nothing. One specific camera or lighting choice does everything.
Forgetting audio — an undescribed soundtrack defaults to generic ambience.
Ignoring version differences — 3.1 follows multi-part structure and audio cues more faithfully; if you're on 3.1, use the Veo 3.1 generator defaults.

Put it together

Start from a copyable example, swap in your subject, and adjust one dial at a time — camera first, then lighting, then audio. That's the fastest feedback loop. When you're ready, open the generator and write your own.