Ijharul Islam
Google Veo 3 is an AI video model from Google DeepMind that turns a text prompt or an image into a short video clip with sound, including dialogue, sound effects and background noise. Its current version, Veo 3.1, makes clips of 4, 6 or 8 seconds in 16:9 or 9:16, and you can use it in Google's Flow and Gemini app, through the Gemini API, or inside tools such as QuickVeed.
This guide explains what Veo 3 and Veo 3.1 are, the exact specs and prices from Google's documentation, the four ways to use it, and how to write prompts that get a usable clip on the first or second try.
The short answer
Veo 3 is Google DeepMind's text-to-video and image-to-video model. It generates video and audio together, so a character can speak a line with matching lip movement. Veo 3.1, released in October 2025, makes 4, 6 or 8 second clips at 720p, 1080p or 2160p in 16:9 or 9:16. You can use it in Flow, the Gemini app, the Gemini API or third-party tools.
Key takeaways
- Veo 3.1 is the current Veo; Google's API docs mark Veo 3 and Veo 3 Fast as deprecated.
- One clip is 4, 6 or 8 seconds at 24 fps, always with sound. 1080p and 2160p need the 8-second length.
- There are three versions: Veo 3.1, the cheaper Veo 3.1 Fast and the lighter Veo 3.1 Lite.
- On the Gemini API, Veo 3.1 costs $0.40 a second at 720p or 1080p, and there is no free tier.
- In QuickVeed, an 8-second clip with audio costs 800 credits on Veo 3.1 and 300 on Veo 3.1 Fast.
Disclosure: we make QuickVeed and show how to do this in it, along with other ways. Details about other products come from their public pages, checked in October 2026 and linked at the end.
What is Google Veo 3?
Veo is Google DeepMind's family of video generation models. You describe a shot in words, or give it a picture, and it returns a video clip. Veo 3 was the version that made the family widely known, because it generated sound in the same pass as the picture: footsteps that land when the feet do, rain you can hear, and people who speak lines you wrote in the prompt.
Veo 3.1 followed on October 15, 2025. Google's announcement lists richer audio, more narrative control, more realistic textures, closer prompt following and better results when you start from an image. Audio also came to Flow's Ingredients to Video, Frames to Video and Extend features at the same time. Today, when people search for Veo 3, the model they can actually use is almost always Veo 3.1.
Google's API documentation lists three current versions. They share the same basic behaviour but differ in features and price:
| Version | What it is for | What it lacks |
|---|---|---|
| Veo 3.1 | Best quality, all features | Nothing in the family; highest price |
| Veo 3.1 Fast | Same features, tuned for speed and cost | Some fine detail compared with full Veo 3.1 |
| Veo 3.1 Lite | Cheapest drafts and simple clips | No 2160p, no reference images, no extension |
A newer Google model, Gemini Omni, arrived in May 2026. Google's Gemini app help page now calls it the app's latest video model, so if you generate a video inside the Gemini app you may be using Omni rather than Veo. Veo 3.1 remains available through Flow, the Gemini API and Vertex AI, and in third-party tools.
Veo 3.1 specs: length, resolution and sound
These figures come from Google's Gemini API documentation for Veo 3.1, checked in October 2026. Other apps that run Veo may offer a subset of them.
| Spec | Veo 3.1 and Veo 3.1 Fast | Veo 3.1 Lite |
|---|---|---|
| Clip length | 4, 6 or 8 seconds | 4, 6 or 8 seconds |
| Resolution | 720p, 1080p, 2160p (1080p and 2160p at 8 s only) | 720p, 1080p (1080p at 8 s only) |
| Aspect ratio | 16:9 or 9:16 | 16:9 or 9:16 |
| Frame rate | 24 fps | 24 fps |
| Sound | Always on: dialogue, effects, ambience | Always on |
| Inputs | Text, image, video (for extension) | Text, image |
| Reference images | Up to 3, to keep a person or product consistent | Not supported |
| First and last frame | Yes | Yes |
| Extension | 7 s per step, up to 20 times, up to 148 s total (720p) | Not supported |
| Watermark | SynthID invisible watermark on every video | SynthID |
Two details catch people out. First, the extension feature only works on Veo-made clips at 720p, so a 2160p hero shot can't be extended. Second, the API keeps generated videos for only two days before deleting them, so download what you want to keep.
Square, 4:3 or 21:9 output isn't listed for Veo, so if you need those shapes you crop afterwards or pick a model that supports them natively. For more on clip lengths across models, see how long AI generated videos can be.
How to use Veo 3: four ways
You can't download Veo and run it yourself. You reach it through one of Google's products or through a tool that pays for API access and passes it on. Here is how the options compare:
| Where | Who it suits | What you pay |
|---|---|---|
| Flow (Google's filmmaking tool) | Creators who want Ingredients, Frames to Video and Extend in one place | Needs a Google AI plan |
| Gemini app | Quick one-off clips from a chat | Needs a Google AI plan; the app's newest video model is Gemini Omni |
| Gemini API or Vertex AI | Developers building Veo into an app or workflow | Per second of video: $0.40 for Veo 3.1, $0.10-$0.12 for Fast at 720p/1080p, $0.05-$0.08 for Lite; no free tier |
| Third-party tools such as QuickVeed | Businesses that want several models in one account without code | The tool's own credits; QuickVeed shows the price before you generate |
On the API, an 8-second Veo 3.1 clip at 1080p works out at $3.20, an 8-second Veo 3.1 Fast clip at 1080p at $0.96, and 2160p costs more ($0.60 a second on Veo 3.1). That is the cheapest way to run Veo if you can write code and handle billing and storage yourself. Most marketers and small teams use Flow or a third-party tool instead, so they can try prompts without building anything.
Using Veo 3.1 in Flow
Flow is Google's tool built around Veo. Its three main ways to make a clip are Ingredients to Video (several reference images guide the characters, objects and style), Frames to Video (you give a start and an end image and Veo makes the motion between them) and Extend (Veo continues a clip from its last second, which Google says can build shots of a minute or more). Google describes the audio and editing features as experimental, so expect some trial and error.
How to use Veo 3.1 in QuickVeed
QuickVeed runs Veo 3.1 and Veo 3.1 Fast alongside 13 other models, so you can try the same prompt on Veo and a cheaper model from one screen.
- Open the AI video generator (in the app it is under Create video).
- Type a prompt, or upload an image to use as the starting frame.
- Choose a model. Clip lengths run from 2 to 15 seconds depending on the model: Veo 3.1 gives 4, 6 or 8 s, Kling v3 3 to 15 s, Seedance 2.0 4 to 15 s, Runway Gen-4.5 5 or 10 s.
- Pick the aspect ratio (16:9, 9:16, 1:1, 4:3, 3:4 or 21:9 where the model supports it) and the resolution (480p to 1080p).
- Check the price shown, then generate. Failed generations are refunded, and every result is saved in your asset library.
Choose Veo 3.1 or Veo 3.1 Fast in step 3 and pick 4, 6 or 8 seconds. Sound is included. Typical prices:
| Example clip in QuickVeed | Credits |
|---|---|
| Vidu Q3 Turbo, 2 s, 540p | 20 |
| Luma Ray 3.2, 5 s, 540p | 38 |
| Kling v3, 5 s, 720p | 210 |
| Veo 3.1 Fast, 8 s with audio | 300 |
| Veo 3.1, 8 s with audio | 800 |
What QuickVeed doesn't do with Veo: it outputs up to 1080p (not 2160p), and it doesn't offer Flow's Extend or Insert tools. For a longer sequence you generate several clips and join them in an editor. If you're choosing between Veo and Kling, our Veo 3.1 vs Kling v3 comparison runs through the trade-offs.
A sensible routine: draft the idea on a cheap model such as Vidu Q3 Turbo or Luma Ray 3.2, move it to Veo 3.1 Fast once the composition works, and spend full Veo 3.1 credits only on the shot that carries the video.
How to write a Veo 3 prompt that works
Google's Veo prompt guide breaks a prompt into subject, action, scene, camera, lens, style and audio. Write it like a short director's note, one shot per prompt, with sound in its own sentence.
| Goal | Example prompt |
|---|---|
| Character speaking | Medium shot of a baker in a flour-dusted apron behind a wooden counter, morning light from the window. She looks at the camera and says: “Everything here was baked before sunrise.” Quiet bakery ambience, an oven door closing. |
| Product shot | Slow orbit around a glass perfume bottle on black marble, soft rim light, shallow depth of field, photorealistic. A low, warm synth tone. |
| Establishing shot | Aerial shot gliding over a coastal town at golden hour, white houses and fishing boats, gentle waves. Seagulls and distant harbour sounds. |
- Keep spoken lines to one sentence. Eight seconds holds roughly 12 to 18 words.
- Name who speaks (“the baker says:”) so Veo knows which mouth to move.
- Describe what you want rather than writing “no” or “don't”; Google's guide recommends this.
- Use camera words Veo understands: close-up, wide shot, slow push in, orbit, tracking shot, low angle.
- For image-to-video, describe only the motion and the sound, not what the picture already shows.
More formulas and twelve copy-ready examples are in how to write prompts for AI video generators.
What Veo 3.1 is good at, and where it falls short
- Good: short scenes with a spoken line, because sound and lip movement come from the same generation.
- Good: following detailed camera and lighting language.
- Good: controlled reveals using a first and last frame.
- Weaker: length. Eight seconds is the cap per generation; Kling v3 and Seedance 2.0 go to 15.
- Weaker: shapes. Only 16:9 and 9:16 are listed.
- Weaker: cost. Full Veo 3.1 is one of the most expensive models per second, in Google's own pricing and in QuickVeed.
Whatever model you use, check hands, text on signs and logos, and any spoken words before you publish. If a clip shows realistic people or events that didn't happen, platforms such as YouTube ask you to disclose that it is altered or synthetic content.
Frequently asked questions
Is Google Veo 3 free?
Not really. The Gemini API lists no free tier for Veo 3.1, and Flow and the Gemini app need a Google AI plan. Some Google products, such as YouTube Shorts, include AI video tools at no cost, and third-party tools like QuickVeed have trials (75 credits).
What is the difference between Veo 3 and Veo 3.1?
Veo 3.1 is the newer version, released in October 2025, with richer audio, better prompt following, more realistic textures and better image-to-video. Google's API now lists Veo 3 as deprecated.
How long can a Veo 3 video be?
One generation is 4, 6 or 8 seconds. Through the API or Flow you can extend a Veo clip in 7-second steps, up to 148 seconds in total on the API.
Does Veo 3 make videos with sound?
Yes. Audio is always on in Veo 3.1: dialogue, sound effects and ambient sound are generated with the picture.
Does Veo 3 add a watermark?
Every Veo video carries SynthID, an invisible watermark that Google's tools can detect. Some apps also add a visible label.
Try it with your own video
Try one prompt on Veo 3.1 Fast and a cheaper model side by side. Type a one-shot prompt with a spoken line, pick Veo 3.1 Fast, and judge the clip and its sound for yourself. QuickVeed plans start at $18/month, and new accounts get a 14-day trial with 75 credits, no card needed. Sign up or see pricing.
Sources
Competitor details in this post were checked against these public pages in October 2026. Products and prices change, so confirm with each vendor before you decide.
- Google Gemini API: Veo 3.1 documentation (lengths, resolutions, extension, SynthID)
- Google blog: Introducing Veo 3.1 and advanced capabilities in Flow (Oct 15, 2025)
- Google Gemini API pricing (Veo 3.1 per-second prices, no free tier)
- Gemini Apps Help: generate videos (Gemini Omni, Google AI plan needed)
- MobileSyrup: Google launches Gemini Omni (free in YouTube Shorts and YouTube Create, May 2026)
- Google Cloud: Veo prompt guide
- YouTube Help: disclosing altered or synthetic content
