How to Use Veo: A Beginner's Guide to Your First AI Video Clip
CodexFuse Team · 2026-10-29T06:07:02.608Z · 12 min read min read
Learn how to use Veo in 2026: pick the right access route, write a prompt that works, and budget your free credits before generating your first 8-second clip.
Learning how to use Veo takes less than five minutes, but using it well takes one honest decision up front: pick the right access route before you touch a prompt. Veo is Google's text-to-video and image-to-video model — you type or upload a starting point, and it returns a short video clip with native audio. It lives inside four different Google surfaces (Flow, the Gemini app, Google Vids, and the developer APIs), and each one has its own models, limits, and credit system. This guide walks you through the whole beginner system: choosing a route, writing a prompt that works, and spending your free credits before they expire.
What is Veo, and where do you actually use it?
Veo is Google's generative video model: it turns a written prompt or a starting image into a short video clip with synchronized sound. The most important thing to know as a beginner is that you never "open Veo" by itself — you reach it through one of four Google apps, and they are not interchangeable. If you are generating inside the Google Gemini beginner guide app, check which model your settings name before following any old "open Gemini and select Video" tutorial, because Google's current video guide describes a model called Gemini Omni there, and instructions written for Veo may not apply to it.
The four access routes are: Google Flow (labs.google/flow), Google's AI filmmaking workspace with an explicit Veo model selector — the best starting point for a beginner; the Gemini app, the casual route for quick single clips; Google Vids (vids.new), the editor that bakes Veo clips into full video projects; and the Gemini API / Vertex AI, the developer route with per-second billing. A beginner needs no API key and no code: start in Flow or Vids.
Veo is 18+ everywhere, and its clips currently cap at 8 seconds. That one constraint shapes the entire workflow: anything longer than 8 seconds is built by chaining clips or extending them, never by writing a longer prompt.
Which access route should you pick as a beginner?
Pick the route that matches the job, not the hype. For a visual project where you want to choose the exact model and settings, use Google Flow. If you want the AI clip inside a real editing timeline with stock media, voiceover, and sharing, use Google Vids. If you just want one quick clip from your phone, use the Gemini app. Use the API only when you need automation and can manage billing and model changes yourself.
| Route | Best for | What a beginner should know |
|---|---|---|
| Google Flow | Choosing the Veo model yourself | Explicit selector: Veo 3.1 Lite, Fast, Quality + newer Omni models; credit cost per clip |
| Google Vids | Full videos with AI clips inside | 10 free Veo generations/month on a free Google account; editor handles the rest |
| Gemini app | Fastest single clip | Current video feature is branded Gemini Omni; paid plans required |
| Gemini API / Vertex | Automation and batch work | Per-second billing, no free tier for Veo; billing managed in Google Cloud |
How do you make your first Veo clip?
Definition: Making a Veo clip means opening Google Flow, choosing a Veo model, writing one structured shot description, and generating an 8-second video from it. Follow these steps once, and you have repeated the whole beginner loop:
- Open labs.google/flow in Chrome or Edge and sign in with a Google account (18+ required).
- Start a new project and switch the prompt box to Video.
- Pick a Veo model: Lite for free drafts, Fast for everyday clips, Quality only for shots you will publish.
- Write your prompt as one shot using the formula below (subject + action + camera + lighting + audio).
- Set the aspect ratio: 16:9 for YouTube, 9:16 for Shorts, Reels, and TikTok.
- Generate once, then watch the result before touching the prompt again.
- Diagnose the miss: wrong subject look? Add a reference image. Wrong movement? Rewrite the camera line only.
- Download the winner. If it needs polish, upscale (subscribers only) or cut it into an edit in the CapCut beginner guide workflow.
What is the Veo prompt formula that actually works?
The prompt formula is a stacked description with five layers: subject, action, camera, lighting, and audio. Veo responds to directorial language, not conversational wishes — "a nice video of a dog" fails; "low-angle tracking shot of a golden retriever running through autumn leaves, tail wagging, golden-hour sunlight, [SFX: leaves crunching]" works because every layer gives the model a decision instead of a guess.
| Layer | What to write | Example |
|---|---|---|
| Subject | Exactly who or what, with visible detail | a golden retriever puppy |
| Action | One clear motion, with a force vector | running through autumn leaves, tail wagging |
| Camera | Shot type + movement + lens feel | low-angle tracking shot, smooth follow |
| Lighting | Time of day + light quality | golden-hour sunlight, soft long shadows |
| Audio | Dialogue or tagged sound effects | [SFX: leaves crunching, distant dog bark] |
Two details separate working prompts from mushy ones. First, refine one layer at a time — if the movement is wrong but the subject is right, rewrite only the camera line. Second, if your character's face drifts between clips, stop describing the face in text and upload a reference image in Ingredients Mode (Flow supports up to three reference images) to lock the identity. Text-only character prompts are the main cause of morphing faces.
How long is a Veo clip, and how do you make longer videos?
A Veo clip is 4, 6, or 8 seconds long — 8 seconds is the maximum for a single generation, and no prompt gets you past it. Longer videos are assembled, not prompted: Flow offers Extend (also called scene extension), which uses the last frame of your clip as the starting point of a new generation of roughly 7 more seconds, and you can chain these to build sequences well past a minute. On the developer API, extensions can stack up to about 148 seconds at 720p.
The beginner trap is generating one clip, disliking it, and regenerating the whole thing ten times. The cheaper move: generate the opening 8 seconds, extend it, and if a later segment fails, extend again from the last good tail frame. Each extension costs another generation's credits, so plan the story beats before you spend.
How does audio work in Veo?
Veo generates native synced audio — sound effects, ambience, and even dialogue — in the same generation as the picture, so most clips need no post-production sound. Write sound into the prompt with bracketed tags: [SFX: ...] for effects, [Ambient: ...] for room tone and background, and quoted dialogue for lip-synced speech (for example: [New arrival, quiet and certain]: "I know."). Keep the audio simple — one dominant sound plus ambience per clip — because a crowded sound list muddies the result.
How much does Veo cost, and how do you use it for free?
Veo is free to try but expensive to spray: every Google account gets 50 Flow credits a day, and each 8-second clip costs a fixed number of credits depending on the model. Plan your first prompt on paper, because the free allowance does not stack and resets daily.
| Model (8s clip) | Credits per clip | Clips from 50 free daily credits |
|---|---|---|
| Veo 3.1 Lite | 10 | 5 |
| Veo 3.1 Fast | 20 | 2 (plus 1 Lite) |
| Veo 3.1 Quality | 100 | 0 — one clip costs double the daily allowance |
That table is the free-credit line the whole beginner strategy hangs on: a free account buys five Lite clips a day, or two Fast clips plus one Lite — but never a single Quality clip, and never an 1080p upscale (upscaling is subscribers-only). Non-subscribers can also be blocked from generating at peak hours. The second free path is Google Vids, which gives every Google account 10 free Veo 3.1 generations per month at 720p inside the editor — capped at 8 seconds each, and chained in the Vids timeline for longer videos.
When the free routes stop being enough, the paid layer is the Google AI plans: AI Plus ($4.99/month) adds 200 Flow credits a month, AI Pro ($19.99/month) adds 1,000, and AI Ultra (from $99.99/month) adds 10,000 to 25,000 — all on top of the daily 50. Monthly credits do not roll over, Flow's default request can generate two videos per click (each charged), and the Gemini API bills per second with no free tier for Veo. The practical rule: draft everything on Lite or Fast, and spend Quality credits only on the clips you will actually publish.
One more cost note for the Vids branch: the free 10 generations a month is the month's budget, not a daily refill — if you want the full editor workflow, read the Google Vids beginner guide before you spend them.
Veo vs Sora: which should a beginner learn first?
For a beginner, the answer is about access, not quality: Veo is the one with a real free path today — 50 daily Flow credits and 10 monthly Vids generations for any Google account — while Sora's free availability has narrowed (OpenAI shut its standalone Sora video tool in 2026, per WinBuzzer's April 2026 reporting). If you already live in Google's ecosystem, Veo is the cheaper skill to build; if you want to compare the models' personalities first, the Sora explainer covers that lane.
The deeper point: both models are 8-second-clip machines, so the transferable skill is the system — the prompt formula, Ingredients-style reference control, and the extend-and-chain workflow — not the brand. Learn the system on Veo's free credits, and it travels.
And if you are cross-shopping AI video tools generally, the Runway beginner guide covers the other major video-model lane with the same beginner structure.
Sources
- DigitalTatva, "Google Veo: How to Use Flow to Make AI Videos (2026)" — checked Oct 12, 2026: Flow credit costs per 8s clip (Lite 10, Fast 20, Quality 100), 50 free daily credits, Google AI plan credit allowances (Plus 200, Pro 1,000, Ultra 10,000–25,000), 18+ and 140+ countries — fetched 2026-10-29.
- AITool.business, "What Is Veo AI? Google Veo 3 Features, Pricing and Review" — 8-second max generation, 4K upscaling 50 credits on Ultra, SynthID in every frame (invisible; visible watermark in India, South Korea, Vietnam), up to three reference images, extension ~7s steps — fetched 2026-10-29.
- Veo3AI.io, "Is Google Flow Free? Pricing, Credits & Limits (2026)" — 50 daily credits buy five Lite or two Fast + one Lite clips; Quality never fits free; no 1080p upscale for non-subscribers; peak-hour blocks for free users; credits do not roll over — fetched 2026-10-29.
- AI Alleyway, "Google Veo 3 Pricing: What It Really Costs Per Clip" — per-clip dollar math on Pro (~$0.20 Lite, ~$0.40 Fast, ~$2.00 Quality), Ultra's cheaper per-credit rate — fetched 2026-10-29.
- ExplainX.ai, "Generate AI Videos in Google Docs with Vids & Veo 3.1" — 10 free Veo 3.1 generations/month in Vids, 720p, 8 seconds each, chaining scenes in the editor — fetched 2026-10-29.
- OnDigitals, "How to use Google Veo 3: Pick the right access route" — checked Oct 12, 2026: Flow vs Gemini app vs editor vs API are not interchangeable; Gemini app video now described as Gemini Omni; record the model shown in your own settings — fetched 2026-10-29.
- AtlasCloud.ai, "How to Use Veo 3.1: Complete AI Video Guide & Prompt Formula (2026)" — 7-layer prompt structure, Ingredients Mode for character consistency, [SFX:]/[Ambient:] audio tags, tail-frame extension — fetched 2026-10-29.
- WinBuzzer, "Google Vids Adds Free AI Video, Music, and Avatars" — April 3, 2026: 10 free Veo 3.1 generations/month for all Google accounts, 720p 8-second clips, Sora standalone tool shut down — fetched 2026-10-29.
- Hooked.so, "Veo 3 Pricing 2026: Plans and Real Cost per Video" — monthly credits don't roll over, daily 50 resets, 2-video default requests double the charge, peak-hour blocks for non-subscribers, Gemini API has no Veo free tier — fetched 2026-10-29.
Frequently Asked Questions
Can I use Veo 3.1 for free?
Yes. Every Google account gets 50 Flow credits a day — enough for five Veo 3.1 Lite clips or two Fast clips — plus 10 free Veo 3.1 generations a month inside Google Vids at 720p. The free path can't produce a Quality clip or an 1080p upscale, and free users may be blocked at peak hours.
How long can a Veo video be?
One generation tops out at 8 seconds. Use Flow's Extend feature to continue from the last frame (~7 more seconds per step) and chain clips together, or build the timeline in Google Vids, which chains scenes without a hard total-length cap.
What is the best prompt formula for Veo?
Stack five layers: subject, action, camera, lighting, and audio — for example: "low-angle tracking shot of a golden retriever running through autumn leaves, golden-hour sunlight, [SFX: leaves crunching]." Refine one layer at a time, and lock characters with reference images in Ingredients Mode instead of describing faces in text.
Does Veo add a watermark?
Every Veo clip carries SynthID, Google's invisible AI-generated-content watermark in each frame. A visible watermark is added automatically in India, South Korea, and Vietnam; in other regions it is optional.
Is Veo better than Sora for beginners?
Veo is the better first skill because it has a real free path — daily Flow credits and monthly Vids generations — while Sora's free access has narrowed. The core beginner system (prompt formula, reference images, extend-and-chain) transfers between both models.