The first trap in any Veo 3.1 tutorial is the name. “Veo” also belongs to a sports camera company, and on some queries that company outranks Google entirely. Everything here is about Google’s Veo 3.1 and Flow.
The second trap is numbers. Tutorials circulate clip lengths and quality settings that no Google page states. This one sticks to what the official pages print, and says so when they print nothing.
Where Veo 3.1 runs
Google’s model page lists six entry points: the Gemini app, Flow, Google Vids, Google AI Studio, the Gemini API and the Gemini Enterprise Agent Platform. The January 2026 update adds YouTube Shorts and the YouTube Create app to that list for the first time.
That list is the whole official answer, and it is worth reading before a tutorial sells you an account. Flow is the filmmaking surface where the creative tools live; the API and Vertex AI are the routes for code.
Write one shot, not a storyboard
Veo generates eight seconds. Google states it flatly — “Veo videos are 8 seconds long” — and the practical consequence is that a prompt describing three unrelated events gets one of them.
One action, one camera idea, one place. If the idea needs a second beat, that is a second generation, not a longer sentence. The getting started guide takes that first clip end to end.
The seven elements Google names
The prompt guide lists seven elements in order: shot framing and motion, style, lighting, character descriptions, location, action, dialogue. Its instruction is to mix them, not to fill all of them in, and its framing sentence is the method: “The more detail you add, the more control you’ll have over the final output.”
Two of the seven carry more weight than the others. Character descriptions reward specificity — the guide’s own example contrasts “a woman in her twenties with wavy brown hair and light freckles” with “a brown-haired woman.” And dialogue is the element Google handles differently from the rest: you can “either pick a topic for characters to talk about, or give characters specific things to say.”
A blank version of that structure is what the two examples below fill in:
Framing and motion: ...
Style: ...
Lighting: ...
Character: ...
Location: ...
Action: ...
Dialogue: ...
Audio: ...One action per beat
For fast, busy footage the guide offers a harder line: “For ultimate control over fast-paced scenes, leave nothing to chance. You can direct every element of the shot with highly-detailed prompts.”
That is not a contradiction of the one-action rule. Detail describes the frame; the action still happens once. The difference between the two is that a detailed prompt says how the mud sprays, not that something else also happens afterwards.
Which settings are published
Three settings have an official sentence behind them. Output “in 1080p and 4K.” Native 9:16 from Ingredients to Video. Eight seconds per generation.
The resolution claim comes with a limit: “upscaling to 1080p and 4K is only available in Flow, the Gemini API and Vertex AI.” The price guide covers what the surfaces charge and what they do not.
Everything else you will read — duration tiers, quality modes, a free generation allowance — appears on vendor and reseller pages rather than Google’s. Those pages may describe their own product accurately, and their numbers are still theirs. The prompt guide also publishes no word limit, no negative-prompt syntax and no troubleshooting section.
Two worked examples
Both prompts are written for eight seconds and use the element order above, because that order is the one Google publishes.
Example one: a product beat
Framing and motion: close-up, static camera, slight handheld drift
Style: documentary product footage, 50mm
Lighting: hard key from the right, deep shadow behind
Character: two hands only, no face in frame
Location: a stainless steel worktop
Action: the lid turns, releases, and is set down on the steel
Dialogue: none
Audio: a rubber seal breaking once, metal touching steel, quiet kitchen behindExample two: a vertical beat
Framing and motion: vertical 9:16, medium shot, slow push-in
Style: editorial fashion film, soft grain
Lighting: window light from frame left, no fill
Character: a courier in a wet orange shell, hood up
Location: a covered walkway, rain beyond the pillars
Action: she checks a parcel label, then steps out into the rain
Dialogue: none
Audio: rain on a roof, one set of footsteps, a van idling somewhere behindNeither prompt has been run, and neither comes with a claimed result. What they demonstrate is the order and the level of detail, which is all the official guidance actually provides.
How to iterate cheaply
The prompt guide has no troubleshooting chapter, so the honest advice is narrow.
Change one variable at a time
Google’s own closing suggestion is the one iteration tool it names: “you can also use Gemini to help you expand on the prompt and include more detail.” That is expansion, not repair — useful when a shot is thin, useless when the camera move is simply wrong.
When a clip misses, decide first whether the problem is in the picture or in the prompt. A wrong subject is a prompt problem. A subject that will not hold still is a length problem, since eight seconds is what you have. Sound is a separate question again, and the audio prompting guide covers it.
When the prompt is not in English
Google publishes no dialogue language list and no prompt-language guidance. The seven element names are English labels from an English page, so any translation of them — including the ones used above — is a reading aid, not official terminology.
What is published is narrower and more useful. Audio “can be integrated into your prompt, or included in a separate section,” which means the structure survives translation. We have not measured whether any one language produces better results, and no official page claims it does.
Frequently asked questions
How long is a Veo 3.1 clip? Eight seconds is the length Google publishes. Everything longer comes from Extend, which continues from the final second of the previous clip.
Do I need a paid plan? Google’s model page lists the surfaces without pricing them. Check the current terms on the surface you intend to use rather than a tutorial’s summary.
Can I write the prompt in my own language? Nothing published forbids it, and the element structure translates. Google publishes no language list.
Does Veo generate music? Google names sound effects, ambient noise and dialogue. Music is not among them, and it is not one of the seven elements.
Write one shot, name the elements you care about, and check each claim against the page it came from.