The new generative AI video model from Google is now available.

Veo, Google's latest generative video model, is now available for companies to incorporate into their content creation processes. This model was first introduced in May, three months after OpenAI showcased its competing product, Sora. Veo has managed to get ahead in the market by launching in a private preview through Google’s Vertex AI platform.

The model is capable of generating 1080p “high-quality” videos in a variety of visual and cinematic styles from text or image-based prompts. Upon its initial announcement, it was mentioned that the generated clips could last over a minute; however, Google has not specified any length restrictions for this preview version. Some of the new clip examples presented in Google's announcement are comparable to what has already been seen from Veo, making it very difficult to distinguish that these videos are AI-generated without close observation.

One of the most impressive examples of Veo clips features a dog, where one can notice how the pattern of its fur and collar remain consistent during its movement. Google's latest text-to-image generator, Imagen 3, will also be available to all Google Cloud customers through Vertex “starting next week,” expanding its initial release in the U.S. that took place in August. Users included in Google’s access list will be able to enjoy new features such as prompt-based photo editing and the ability to “infuse your own brand, style, logo, subject, or product features” into the generated images.

However, Veo is not perfect, as seen in a generated video of a concert where light shines through a person's hand in the upper left corner. Google assures that Veo and Imagen 3 have built-in safeguards to prevent the generation of harmful content or copyright violations, although the latter has been found to be easily circumvented. Additionally, everything produced by Veo and Imagen 3 is embedded with DeepMind's SynthID technology, a type of invisible digital watermark that, according to Google, can “reduce misinformation and concerns about misattribution.” This concept is similar to Adobe's Content Credentials system, which can be embedded in content produced by its own generative image and video models.

With Google's video model now available, OpenAI is notably lagging behind its competitors and is running out of time to fulfill its promise of launching Sora before the end of 2024. The emergence of AI-generated content in advertisements, such as Coca-Cola's recent holiday campaign, indicates that companies are not willing to wait for Sora; in fact, according to Google, 86 percent of organizations already using generative AI are seeing an increase in their revenue.