Glossary
What is Google Veo?
In short
Google Veo is a Google DeepMind model that creates short videos with sound from a text description or an image.
Models and products · Veo
MMOOZZGG.com editorsChecked: 1 min read
For a child
Imagine a magic projector. You whisper to it, “a kitten plays with a ball of yarn in the rain,” and a small film appears on the wall, with the sound of raindrops and purring. The film is very short, like one scene from a cartoon. Sometimes the projector mixes up details, and the kitten may get an extra paw. So adults look at the result before they show it to others.
In plain words
It is a video generator from Google. You describe a scene in words or upload a picture, and the model creates a clip of a few seconds with sound included: street noise, footsteps, and the lines of the characters. You can set the first and the last frame to get a smooth transition, or extend a finished scene. Google showed the first version on 14 May 2024, and the current line in the documentation is Veo 3.1. Example: a shop describes a sneaker that slowly turns on a stand under soft light and gets an intro for its product page.
For business
Marketing teams use such models for rough clips, intros, and ad variants before a shoot. The cost depends on the length, the resolution, and the model version, so check Google’s price list at the time you launch. There is no official access from Russia: the country is not on the list of regions for the Gemini API and Google AI Studio, nor among the countries where the Gemini app works (checked 16 October 2026). Google does not publish the weights of Veo, so you cannot run the model on your own server. Before you publish clips, check the rights to people and brands in the frame and the rules for labeling generated content. Plan how to build generation into a regular video schedule before you launch a channel.
For a specialist
Veo is a video generation model (text-to-video, image-to-video) with native sound: the audio track is created together with the image, not added later. As of 16 October 2026, the Gemini API offers Veo 3.1, Veo 3.1 Fast, and Veo 3.1 Lite as preview versions, and Veo 3 is marked deprecated. A clip lasts 4, 6, or 8 seconds at 24 frames per second, in 16:9 or 9:16, at 720p or 1080p, and the two senior versions also offer 4K. The same two versions accept up to three reference images and a first and last frame, and they can extend a clip (video extension). The finished video stays on the server for two days and is then deleted. Each clip carries a SynthID watermark, which you can check on the SynthID verification platform. When Google announced Veo in May 2024, it promised 1080p videos longer than a minute.
Example
An agency prepares ad variants for a coffee shop. From a description, the model gives eight-second clips with sound, and the director picks a good one and plans the real shoot from it.
Related terms
Questions and answers
Sources
- Google, Generate videos with Veo in the Gemini API, checked 16 October 2026ai.google.dev
- Google DeepMind, Veo, checked 16 October 2026deepmind.google
- Google, Generating video and images with Veo and Imagen 3, 14.05.2024blog.google
- Google, Gemini API: Available regions, checked 16 October 2026ai.google.dev
- Google, Where you can use the Gemini web app, checked 16 October 2026support.google.com
Not sure whether your company needs this?
AI solutions map