Industry reports or surveys on AI video generation usage

Google DeepMind's Veo 2 Promises 4K Video Clips, Enhanced Realism

Dec 16, 2024, 05:24 PM

Google DeepMind has unveiled Veo 2, a next-generation AI model for video generation, which promises to create high-quality video clips up to 4K resolution and over two minutes in length. This model is designed to rival OpenAI's Sora, claiming superior performance in terms of video quality and adherence to user prompts. Veo 2 introduces improved understanding of real-world physics, human movement, and expression, as well as enhanced camera control capabilities for better positioning and movement. It also features a reduced frequency of 'hallucinations,' or unwanted details in the generated videos. To ensure authenticity, Veo 2 outputs include an invisible SynthID watermark. Alongside Veo 2, Google announced updates to its image generation model, Imagen 3, which now offers brighter, better-composed images with richer details and textures, and can render more diverse art styles accurately. Both models are being made available through Google Labs' tools, VideoFX for video and ImageFX for images, with a waitlist for access. Google is also expanding the availability of Veo 2 to YouTube Shorts and other products in the coming year. Additionally, Google Labs has introduced a new experiment, Whisk, allowing users to input or create images to visualize and remix ideas.