Veo is Google DeepMind's AI video generator: you describe a scene in words, or hand it an image, and it returns a short video clip with matching sound. The current version, Veo 3.1, makes 4 to 8 second clips at up to 4K, with dialogue, sound effects and music generated in the same pass as the picture.
That one sentence hides a lot. There are four Veo generations, three tiers of the current model, and at least four official places to use it. This guide covers what Veo is, how each version differs, what it does well and badly, whether you can use it for free, and how it stacks up against Sora 2 and the rest.
TL;DR
- What it is: a family of text-to-video and image-to-video models from Google DeepMind, first shown in May 2024.
- Latest version: Veo 3.1 (October 2025, expanded January 2026), available as Veo 3.1, Veo 3.1 Fast and Veo 3.1 Lite.
- Why people care: since Veo 3, every clip comes with native audio, including lip-synced speech.
- Output: 4, 6 or 8 second clips, 720p to 4K, landscape (16:9) or vertical (9:16). Scene extension chains clips into scenes of a minute or more.
- Is it free? Partly. Google and third-party apps such as Veo Studio offer limited free generations; regular use is paid.
- Veo 4: not released as of October 2026.
What is Google Veo?
Google Veo is a generative video model built by Google DeepMind, the AI lab behind Gemini. It belongs to the same family of tools as Imagen (images) and Lyria (music), and it powers video generation across Google's products.
In practice, Veo does two jobs:
- Text to video. You write a prompt such as "a drone shot over a foggy pine forest at sunrise, soft piano music" and Veo renders it.
- Image to video. You upload a still, a product photo, a character or a first and last frame, and Veo animates it.
Under the hood, Veo is a diffusion model. It starts from noise and refines it, step by step, into frames that match your prompt, while keeping motion, lighting and physics consistent from one frame to the next. Since Veo 3, the model generates the soundtrack at the same time, so a door slam lands on the right frame and a character's lips match their line.
What is Veo 3?
Veo 3 is the third generation, launched at Google I/O in May 2025. It was the first mainstream video model to generate synchronized audio natively, and that is what made it go viral: talking characters, street interviews and fake ads that sounded real. If you have searched for "Veo 3", you were most likely looking at clips from this model or its successor.
What is Veo 3.1?
Veo 3.1 is the current model. It keeps Veo 3's audio and adds the controls filmmakers asked for: up to three reference images for consistent characters and products, first and last frame control, scene extension for longer shots, native vertical video and 4K output. Read the full breakdown on our Veo 3.1 page.
One note on names: Veo Studio (useveo.ai) is an independent app that runs the Veo models. It is not made by or affiliated with Google.
Veo versions: from Veo 1 to Veo 3.1 Lite
Veo has moved fast: six releases in under two years, each fixing the biggest complaint about the last one.
| Released | Version | What changed |
|---|---|---|
| May 2024 | Veo | First version, shown at Google I/O 2024. 1080p clips with a grasp of cinematic terms like "timelapse" and "aerial shot". |
| December 2024 | Veo 2 | Sharper detail, better physics and human motion, higher resolution. Reached the Gemini app in April 2025. |
| May 2025 | Veo 3 | Native audio: dialogue, sound effects and ambience generated with the video. Launched alongside Google's Flow filmmaking tool. |
| October 2025 | Veo 3.1 and Veo 3.1 Fast | Richer audio, stronger prompt adherence, reference images ("ingredients to video"), first and last frame, scene extension. |
| January 2026 | Veo 3.1 update | Better character consistency from references, native 9:16 vertical output, 1080p and 4K output. |
| March 2026 | Veo 3.1 Lite | Cheapest tier, as fast as Veo 3.1 Fast, for high-volume work, 720p and 1080p. |
In May 2026 Google also launched Gemini Omni, a separate multimodal model that edits video through conversation. It sits next to Veo rather than replacing it: Veo 3.1 remains Google's dedicated video model and API.
What can Veo do?
Veo 3.1 is built for directed shots rather than random clips. These are the features that matter most in day-to-day use.
Native audio and dialogue
Veo writes the soundtrack with the picture. Put a line in quotes and a character says it, with matching lip movement. Describe sounds ("rain on a tin roof", "distant sirens") and they land in sync. You do not need a separate voice-over or sound design step for most social and ad content.
Text to video and image to video
A prompt alone is enough. Add an image and Veo uses it as the opening frame, or as a look to follow. This is how most product videos are made: the product photo goes in, a 360-degree turntable or lifestyle shot comes out.
Reference images ("ingredients to video")
Give Veo up to three images of a person, object or style and it keeps them consistent across the clip. In Veo Studio you tag them in the prompt, for example "@[image1] wears @[image2]'s jacket". It is the closest thing to casting and wardrobe in AI video.
First and last frame
Set a start image and an end image and Veo generates the motion between them, audio included. It is ideal for transitions, transformations and reveals.
Scene extension
Each generation is short, but you can extend a finished clip. Veo continues from its final second, keeping the world, characters and sound coherent. In Veo Studio each step adds 7 seconds, up to about 148 seconds of continuous video.
Cinematic control
Veo understands film language: dolly in, crane shot, rack focus, shallow depth of field, golden hour, 35mm film grain. The more you direct like a cinematographer, the more control you get. Our Veo 3 prompt guide covers this vocabulary in depth.
Formats built for publishing
Clips come in 16:9 for YouTube and ads, or 9:16 for TikTok, Reels and Shorts, at 720p, 1080p or 4K.
Veo 3.1 vs Veo 3.1 Fast vs Veo 3.1 Lite vs Veo 3
Pick Veo 3.1 for final shots, Fast for drafts and most social content, and Lite when you need volume at the lowest cost. Veo 3 is still available but has been superseded.
| Model | Max resolution | Speed | Native audio | Best for |
|---|---|---|---|---|
| Veo 3.1 | 4K | Slowest | Yes | Hero shots, ads, dialogue scenes, anything final |
| Veo 3.1 Fast | 4K | Fast | Yes | Iterating on ideas, social video, most day-to-day work |
| Veo 3.1 Lite | 1080p | Fast | Yes | High-volume drafts, B-roll, testing prompts cheaply |
| Veo 3 | 1080p | Medium | Yes | Matching older projects made with Veo 3 |
| Veo 3 Fast | 1080p | Fast | Yes | Quick Veo 3-style drafts |
The 3.1 models generate 4, 6 or 8 second clips in 16:9 or 9:16. Advanced controls such as scene extension are tied to the 3.1 models, so start there unless you have a reason not to. In Veo Studio, Veo 3.1 is the default for every account and you can switch models from the prompt box.
Where can you use Veo?
Veo is not a standalone Google app. You reach it through one of these doors:
| Where | Who it suits | How you pay |
|---|---|---|
| Gemini app | Casual users already in Google's ecosystem | Google AI subscription, with daily or monthly limits |
| Google Flow | Filmmakers who want a timeline and scene tools | Monthly Flow credits included in Google AI plans |
| Gemini API and Google AI Studio | Developers building Veo into apps | Per second of video generated |
| Vertex AI (Google Cloud) | Enterprises with Google Cloud contracts | Per second of video generated |
| Veo Studio and other third-party apps | Creators and marketers who want a simple prompt box | Free plan, monthly plans or one-time credit packs |
Developers who want Veo inside their own product can also start with our Veo 3 API page.
Is Veo free?
Veo is free to try but not free to use at scale. Generating video is expensive: on Google's own Gemini API, Veo 3.1 costs between $0.05 and $0.60 per second of video depending on tier and resolution, so a single 8-second clip runs from $0.40 to $4.80.
Your free options:
- Veo Studio free plan. 160 credits a month, enough for up to 2 videos at 720p. No card required and no watermark added. Try Veo 3 free.
- Google trials. Google periodically includes limited Veo generations in Gemini and Flow, and new Google AI subscribers often get a trial month. The details change often, so check Google's current offer.
For regular use, paid Veo Studio plans start at $14 a month for up to 50 videos, with a commercial licence. See pricing for every plan and one-time credit packs that never expire.
Veo vs Sora 2 and other AI video models
Veo 3.1's edge is control: 4K output, native audio, reference images and scene extension in one model. Rivals win on specific things, so the right choice depends on the shot.
| Model | Maker | Where it stands out |
|---|---|---|
| Veo 3.1 | Google DeepMind | Cinematic camera control, lip-synced dialogue, 4K, consistent characters from references, longer scenes via extension |
| Sora 2 | OpenAI | Convincing physics in single clips, longer single generations, a social app built around it |
| Kling | Kuaishou | Dynamic motion and action, good value |
| Seedance | ByteDance | Fast, strong multi-shot storytelling |
| Runway | Runway | Editing tools and a full creative suite around the model |
Veo 3 vs Sora 2: both generate video with sound. Veo tends to win on prompt adherence, camera direction and dialogue scenes; Sora 2 is often preferred for physically complex action in one take. Many professional teams use Veo for finished work and another model for experiments.
The field changes monthly, so treat any comparison, including this one, as a snapshot from October 2026.
Veo's limitations
Veo is impressive, but knowing its weak spots saves credits.
- Short clips. One generation is 8 seconds at most. Longer videos mean extending scenes or editing several clips together.
- Text on screen. Signs, labels and logos can come out garbled. Add text in an editor afterward.
- Complex physics. Fast hand movements, crowds and objects passing through each other can still break.
- Hit rate. Expect to generate two to four versions to get a keeper. Clear, specific prompts raise the odds.
- Content rules. Google's safety filters block violent, explicit and some real-person content, and some prompts are refused without a clear reason.
SynthID: how Veo marks its videos
Every Veo video carries SynthID, an invisible watermark from Google DeepMind embedded in the pixels. It does not change how the video looks, and it lets tools identify the clip as AI-generated. Some Google apps also add a visible watermark; Veo Studio does not add one.
Is there a Veo 4?
No. As of October 2026, Google has not announced Veo 4, and Veo 3.1 is the newest Veo model. Google's newest video work, Gemini Omni, launched in May 2026 as a separate model inside Gemini rather than as a new Veo. We will update this guide when that changes.
Veo FAQ
What is Veo used for?
People use Veo for social videos, product ads, music videos, short films, explainer B-roll and storyboards. Its native audio makes it especially good for anything with dialogue or a voice.
Who made Veo?
Google DeepMind develops Veo. Google offers it through Gemini, Flow, the Gemini API and Vertex AI, and third-party apps like Veo Studio provide access too.
Is Veo 3 free?
You can try it free with limits. Veo Studio's free plan includes up to 2 videos a month at 720p with no card, and Google runs occasional trials. Ongoing use requires a paid plan or API credits.
What is the difference between Veo 3 and Veo 3.1?
Veo 3.1 adds reference images, first and last frame control, scene extension, native vertical video and 4K output, plus better audio and prompt adherence. Veo 3 introduced native audio; Veo 3.1 refines it.
How long can a Veo video be?
A single Veo 3.1 clip is 4, 6 or 8 seconds. With scene extension you can grow one continuous scene to about 148 seconds in Veo Studio.
Does Veo make sound?
Yes. Since Veo 3, music, sound effects, ambience and dialogue are generated together with the video and synced to the action.
Can I use Veo videos commercially?
On paid Veo Studio plans, yes: every paid plan includes a commercial licence. If you use Veo through Google, check the terms of your Google plan.
Try Veo today
The fastest way to understand Veo is to make something with it. Create a free Veo Studio account, type a prompt and get your first clip with sound in minutes. Next, read how to use Veo 3 for a full walkthrough, and the Veo 3 prompt guide for prompts that work.
