Veo 3 made a strange first impression on most people: everyone saw the clips, almost nobody could find the button. Google shipped the model inside several different products — Gemini, Flow, Vertex AI — each with its own access rules, country list, and subscription tier, so "how to use Veo 3" became a genuinely confusing question with five half-answers.
This tutorial gives you the one complete answer. The fastest way to use Veo 3 today is Veo AI: a browser studio where Veo 3.1 Pro, Veo 3.1 Fast, and 40 other models run on pay-as-you-go credits. You sign up, you write a prompt, you generate — and your first clips are covered by free starter credits, so the whole walkthrough below costs nothing to follow along with.
We will go through the eight steps of your first generation, then cover how Veo 3 works inside Google’s own apps for completeness, which prompts to spend your first credits on, and the mistakes that waste a beginner’s first afternoon.
The 8-step Veo 3 walkthrough
Here is the full loop, from no account to a downloaded video. If you have used any web app before, expect the first clip within a few minutes.
1. Sign up in the browser
Go to the sign-up page and create an account — it works in any modern browser on desktop or phone, with nothing to install and no waitlist. New accounts come with free starter credits, which is enough to generate real clips while you learn. There is no subscription to start and none required later; credits are the only currency.
2. Open the Create studio
Once signed in, head to the Create studio. This is the single screen where everything happens: the prompt box, the model selector, the output settings, and your generated clips. Everything in this tutorial takes place here.
3. Pick your model: Veo 3.1 Pro or Fast
In the model selector, choose between Veo 3.1 Pro and Veo 3.1 Fast. Pro is the quality tier — use it for final renders. Fast trades a little fidelity for quicker, cheaper generations — the right choice while you are learning and iterating. Both produce native audio: speech, ambience, and effects generated together with the picture.
4. Write the prompt
A working Veo prompt names four things: the subject (who or what), the action (what they do), the camera (how the shot is framed and moves), and the lighting or mood. For example: "A street vendor flips dumplings in a sizzling wok, slow push-in at counter height, warm lantern light, steam catching the glow." If you want spoken dialogue, put the exact line in quotation marks inside the prompt — Veo lip-syncs quoted speech, and describing the sound you want ("sizzling oil, distant market chatter") shapes the audio track.
5. Choose duration, resolution, and aspect ratio
Set the clip length to 4, 6, or 8 seconds, the resolution to 720p or 1080p, and the aspect ratio to 16:9 for landscape or 9:16 for vertical. A practical default for learning: 8 seconds, 720p, 16:9. Go vertical when the clip is destined for TikTok, Reels, or Shorts, and reserve 1080p for renders you intend to keep.
6. Generate — the cost shows first
Before you commit, the studio displays exactly how many credits this generation will cost with your current settings. No meter running in the background, no surprise at month-end — you see the price, you click generate, and the clip renders in the browser with its audio.
7. Iterate with negative prompts and references
Your first result will be close but rarely perfect — that is normal. Change one variable and re-roll. Negative prompts remove what you keep seeing and do not want ("no text overlays, no extra hands"); reference images lock a face, product, or style so it stays consistent across takes; keyframes let you pin the first and last frame when you need the shot to start and end exactly somewhere.
8. Download — watermark-free
When a take lands, download it. Every export is full quality with no visible watermark — ready to post directly, or to drop into an editor alongside the rest of your project. That is the whole loop; from here, using Veo 3 is just steps 4 through 8 on repeat.
Using Veo 3 inside Google’s own apps
For completeness: Google offers Veo through two consumer surfaces, and both are legitimate routes if you already pay for them. In the Gemini app, you ask Gemini to make a video and it renders a Veo clip in the chat — simple, but sustained video generation sits behind the Google AI Pro and Ultra subscriptions, with monthly allowances and country-by-country availability. Flow, at labs.google, is Google’s filmmaking workspace built on Veo, aimed at multi-shot projects — also subscription-gated.
We keep dedicated, regularly updated guides to both routes: see the Google Flow guide and the Veo in Gemini guide for setup details. The short version of the trade-off: Google’s surfaces make sense if you already hold an AI Pro or Ultra plan; if you do not, pay-as-you-go credits on Veo AI get you the same model family without the monthly commitment.
How to open Veo 3 (the direct answer)
A lot of people searching "how to open Veo 3" just want the door. Veo 3 is a model, not an app, so there is no Veo icon to tap — you open one of the products that runs it. Your three options: veoai.io in any browser (no subscription, credits-based — the route this tutorial uses), gemini.google.com with a Google AI Pro or Ultra plan, or labs.google for Flow with the same plans. If you are outside the US, note that Google’s rollouts are country-dependent, while a browser studio works wherever you are — which is why "how to use Veo 3 in India" and similar searches usually end at the first option.
Your first five prompts: spend starter credits wisely
The fastest way to learn Veo is to generate five deliberately different clips rather than five variations of one idea. Each prompt below teaches you a different control surface of the model.
- A talking subject: a barista looks into the camera and says "First one’s on the house" — teaches you quoted dialogue and lip-sync.
- A camera move: a drone shot rising over a coastal cliff at dawn — teaches camera language and how Veo interprets motion words.
- A lighting study: the same desk scene described once with "harsh noon window light" and once with "single warm lamp at night" — teaches how much lighting words steer the look.
- A vertical clip: a 9:16 street-style shot built for Reels — teaches format thinking before you need it.
- An image-to-video test: upload any photo you own and animate it — teaches the i2v workflow you will use constantly later.
Common beginner mistakes (and the fix for each)
- Cramming three scenes into one prompt — a 4–8 second clip holds one subject and one action; split bigger ideas into multiple generations and cut them together.
- Forgetting the camera — without a framing instruction Veo picks its own, usually a generic drifting wide shot; one phrase ("locked-off close-up", "handheld follow") fixes it.
- Re-rolling without changing anything — same prompt, same settings mostly buys the same weaknesses; change exactly one variable per take so you learn what moved.
- Burning Pro credits on drafts — explore on Veo 3.1 Fast (or the audio-free Veo 3.1 Lite, the cheapest Veo), then re-run the winning prompt on Pro.
- Describing sound nowhere — Veo generates audio whether you direct it or not; an undirected soundtrack is the most common reason a good-looking clip feels off.
- Ignoring negative prompts — if an artifact shows up twice, name it in the negative prompt instead of hoping the next roll skips it.
How does Veo 3 actually work?
You do not need theory to use the model, but a one-paragraph picture helps your prompting. Veo is a diffusion-based video model: it was trained on vast paired examples of footage, sound, and description until it learned the statistical shape of how real scenes look, move, and sound. At generation time it starts from noise and, guided by your prompt, refines that noise step by step into frames and a matching audio track. The practical consequence: the model fills every gap you leave with the most statistically ordinary choice. Specific prompts beat vague ones not because the model rewards effort, but because specificity removes its room to be generic.