The Google I/O 2024 keynote ran for 112 minutes, and the company made several important announcements focused on artificial intelligence (AI). Announcements ranged from new AI models to integrating AI into Google products, but perhaps one of the most interesting introductions was his Veo, an AI-powered video generation model that can generate 1080p resolution videos. . The tech giant said the AI tool can generate videos longer than one minute. Notably, OpenAI also announced a video AI model called “Sora” in February.
During the event, Google DeepMind co-founder and CEO Demis Hassabis announced Veo. Announcing the AI model, he said: “Today, we are excited to announce the newest and most capable generative video model called Veo. Veo creates high-quality 1080p video from text, images, and video prompts. Capture the details of your instructions in a cinematic style.”
The tech giant claims that Veo can follow prompts closely, understand the nuances and tone of a phrase, and generate a video that resembles it. AI models can generate videos in a variety of styles, including time-lapse, close-ups, high-speed tracking shots, aerial shots, and shots with varying lighting and depth of field. Apart from generating the video, the AI model can also edit the video once the user provides the initial video and prompts to add or remove something. Additionally, you can generate videos longer than 1 minute through a single prompt or multiple consecutive prompts.
To solve the problem of consistency in video generation models, Veo uses latent diffusion transformers. This helps reduce instances where characters, objects, or entire scenes flicker, jump, or morph unexpectedly between frames. Google emphasized that videos created by Veo will be watermarked using SynthID, the company's in-house tool for watermarking and identifying AI-generated content. This model will soon be available to some creators via Google Labs' VideoFX tool.
Similarities between Veo and OpenAI's Sora
Although neither AI model is yet publicly available, both have some similarities. Veo can produce his 1080p videos that are over a minute long, while OpenAI's Sora can produce videos up to 60 seconds long. Both models can generate videos from text prompts, images, and videos. Both are based on diffusion models and can generate videos from multiple shots, styles, and cinematography techniques. Sora and Veo also come with AI-generated content labels. Sora uses the Coalition for Content Provenance and Authenticity (C2PA) standard, and Veo uses native SynthID.
