Google Unveils Veo 3.1 with image to dynamic video geneneration feature
Google has introduced Veo 3.1, the latest version of its generative AI video generation model, with a new feature called Ingredients to Video. This update enables users to transform reference images into dynamic, expressive videos through simple prompts, focusing on improved consistency, creativity, and user control.
Veo 3.1 builds on previous versions by generating videos that feel more natural and engaging. One key advancement is in identity consistency, where characters maintain their appearances across different scenes, supporting improved narrative flow.
Similarly, the model retains backgrounds and objects, enabling creators to reuse elements without disruption. This consistency extends to blending various components—such as characters, textures, and stylized backgrounds—into seamless clips.
A notable addition is native support for vertical video formats in a 9:16 aspect ratio, optimized for mobile viewing. This eliminates the need for cropping, making it ideal for platforms like YouTube Shorts.
Veo 3.1 also introduces advanced upscaling capabilities, delivering videos in 1080p for editing purposes and up to 4K for high-end productions requiring sharp textures and clarity. These features cater to both casual users and professionals, with upscaling available in tools like Flow, Gemini API, and Vertex AI.
The process is straightforward: users input ingredient images—potentially generated via Gemini—and add short prompts to create lively videos with enhanced dialogue and storytelling.
For example, the model can produce clips demonstrating character persistence in changing environments or the integration of diverse textures into vibrant scenes. Google embeds SynthID, a digital watermark, in all generated content to promote transparency and allow verification through the Gemini app.
Applications span from everyday creators making shareable content on mobile apps to enterprises using it for tailored workflows in Google Vids. Professionals benefit from the high-fidelity outputs, while casual users can experiment in the Gemini app or YouTube Create.
The 3.1 Veo model is now accessible across Google platforms, including the Gemini app for portrait mode videos and YouTube for short-form content.
(source)