Google's Veo 2 makes a big, faithful leap to 4K compared to the first Veo's 1080p

Google has launched Veo 2, its latest video-generating model, not too long after bringing the first Veo to businesses and enterprises.

The Mountain View tech giant says you can create videos up to 4K in, unlike the first Veo, which had a 1080p resolution cap. It also supports word prompts even for lens types, angles, and effects.

Google boasts that Veo 2 “brings an improved understanding of real-world physics and the nuances of human movement and expression, which helps improve its detail and realism overall.”

AI videos have their problems. You may have come across a lot of AI videos with unrealistic details like people with six fingers because AI video models do tend to struggle with physics.  

And so for that, Google says that Veo 2 minimizes these common AI video issues, like hallucinated details. It also has SynthID watermarks, aligning with Google’s participation in the C2PA initiative, which aims to distinguish AI-generated visuals, a while ago.

Currently available in Google Labs’ VideoFX tool, Veo 2 will expand to YouTube Shorts and other products in 2025. Within the same announcement, Google also said that it’s revamping the Imagen 3 image-generating model and launched the Whisk AI generator, which lets you generate AI images using your own images.

The announcement arrived amid hefty competition from OpenAI and its video model, Sora. Launched for general availability during the Microsoft-backed company’s 12 days of “shipmas,” Sora arrived alongside Sora “Turbo,” its faster variant.

But it has not been an easy launch for the Microsoft-backed company, though. A group of artists, who volunteered to beta-test Sora before its launch, exposed the model’s front end. It was available to try for non-beta testers before OpenAI shut it down. And on top of that, Sora is not yet available for most European users.