Google expands Veo 3.1 in the Gemini API with 4K video

Google has updated Veo 3.1 in the Gemini API and Google AI Studio with more control over video creation, new format options, and resolution up to 4K.

The update is aimed at developers who want to integrate video generation into their applications without having to manually solve problems such as characters changing appearance, inconsistent backgrounds, or poorly framed vertical videos.

More consistent characters and settings

The Ingredients to Video feature lets you use images or reference elements to build a video. With the announced improvement, Veo 3.1 does a better job of preserving character identity and setting details from one scene to the next.

In practice, you can provide images of a character, product, or location and ask for them to appear in different situations. The goal is to keep the character from changing their face or clothes without reason and to maintain a recognizable appearance throughout the video.

That does not mean the result will be identical in every frame, but the model has more information to preserve visual continuity.

Vertical videos without cropping

Veo 3.1 also adds native 9:16 vertical generation within Ingredients to Video. This is the standard format for TikTok, Instagram Reels, and YouTube Shorts.

Until now, a common way to adapt a horizontal video was to crop it afterward. That could cut off faces, products, or important parts of the scene. The new mode generates the entire video vertically, with a composition designed for mobile screens and faster results, according to Google.

For developers, this makes it possible to create social media content directly without adding a separate cropping and reframing step.

Higher resolution: 1080p and 4K

The update improves video clarity at 1080p and adds 4K generation. The first option is designed to deliver sharper images for common internet uses, while the second targets projects that need more detail and a larger screen.

These improvements may be useful for ads, product demonstrations, social media content, and audiovisual prototypes. Even so, final quality will also depend on the input images, instructions, and type of scene being generated.

The new capabilities have been available since January 13, 2026, in the Gemini API and Google AI Studio. Google is also offering them in Vertex AI for businesses and includes SynthID, a digital watermark designed to help identify AI-generated content.

For you, the most relevant change is that creating vertical video, maintaining visual identity, and working at higher resolution are now part of the same development workflow. The next thing to watch is how these features perform in longer scenes with multiple characters, where consistency remains one of the main challenges in AI-generated video.