Google’s Veo 2 gets a major upgrade with new AI-powered features

Google I/O

Google’s Veo 2 video generation model is getting a serious creative boost. As the company continues to develop Veo 3, it’s rolling out a set of powerful new features for Veo 2—designed specifically with filmmakers and visual storytellers in mind.

The headline upgrade? Reference-powered video generation. Creators can now feed Veo images of characters, objects, or even entire scenes and visual styles. The model uses this input to generate consistent, stylized video sequences—bringing a new level of directorial control to AI video creation.

Also new are camera controls. Users can now define complex camera moves like dolly shots, zooms, and rotations, letting them orchestrate cinematic sequences rather than relying on static or generic framing.

There’s also a new outpainting capability that dynamically expands the frame—turning portrait videos into landscape or adjusting aspect ratios to fit any screen, all while adding realistic environmental elements beyond the original frame.

One of the most intriguing updates is object-level editing. Veo 2 can now add or remove objects from scenes, intelligently understanding scale, interaction, and lighting. That means you can clean up a cluttered shot, or insert a key prop, without reshooting—or even shooting at all.

Meanwhile, Google also announced Veo 3 — its latest model at the 2025 I/O event. "Veo 3, our new state-of-the-art video generation model, not only improves on the quality of Veo 2, but for the first time, can also generate videos with audio — traffic noises in the background of a city street scene, birds singing in a park, even dialogue between characters,” said Google in a blog post.