Google DeepMind has launched Veo 2, a next-generation video-generation AI designed to outperform rivals like OpenAI.
Announced on Monday, Veo 2 is the successor to DeepMind’s original Veo model and can create two-minute-plus videos in resolutions up to 4K (4096 x 2160 pixels), making it four times sharper and over six times longer than OpenAI’s Sora can currently produce.
However, these capabilities are yet to be fully realized. In Google’s experimental tool, VideoFX, where Veo 2 is now being tested, video output is capped at 720p resolution and eight seconds in length.
By comparison, Sora can generate clips up to 1080p and 20 seconds long. Access to VideoFX remains limited, but Google plans to expand the user base this week.
VP of product at DeepMind Eli Collins, hinted at broader applications. “Over the coming months, we’ll iterate based on user feedback and integrate Veo 2’s capabilities across the Google ecosystem,” he said.
Veo 2 improves video clarity, physics modeling, and camera controls, enabling more precise angles, fluid dynamics, and cinematic effects. It also supports nuanced human expressions and complex motions, making it a powerful tool for creative and cinematic applications.
