Google launches Veo model to generate video with artificial intelligence for companies
Google announced the availability of its new Veo model for generating video using artificial intelligence for companies, to integrate it into content production processes, in addition to the Imagen 3 model for generating images, in a statement published on its official website.
This launch comes after the initial unveiling of the model last May, and after OpenAI presented its competing model Sora, but Google has now outdone itself by launching the Veo model first via the Vertex AI platform, while the Sora model is still under development.
Veo can create 1080p videos in a variety of visual and cinematic styles using text or image commands.
When the model was announced earlier this year, the clips produced were about a minute long, but Google didn't set video length limits in the beta, and new clips Google has revealed show results that are hard to distinguish from real videos.
A prime example is a video of a dog that displays fine detail, with the fur and collar patterns remaining consistent as it moves, demonstrating the model's strength in producing high-quality clips.
In addition to Veo, Google will make the latest version of its Imagen 3 text-to-image model available to all Google Cloud customers on Vertex starting next week, following its initial launch in the US via the AI Test Kitchen in August.
Google will provide new features such as editing images based on text commands, and adding logos or user-specific elements to the produced images.
The Veo model is not without its flaws, despite its advanced capabilities, such as lighting defects that appear in videos, as happened in one of the test clips, where the light penetrated a person's hand in an unnatural way.
However, Google stresses that Veo and Imagen 3 are equipped with protection mechanisms to prevent the generation of harmful content or intellectual property infringement. Additionally, Google has included DeepMind’s SynthID technology in all generated content, an invisible digital watermark that aims to reduce the risk of misinformation and distinguish content as AI-generated or modified.
With Veo, OpenAI finds itself lagging behind, having yet to deliver on its promise to launch Sora before the end of 2024. As AI-generated content becomes more widely used in advertising campaigns, Google data suggests that 86% of organizations using generative AI have seen revenue increases, giving businesses a huge incentive to adopt Google’s technology without waiting for its competitors.
Google showed a sample promotional video for travel company Agoda, in which it employed new AI models Veo and Imagen 3.