Google's Veo 3.1 Can Turn Separate Images Into A Single Video

Artemisia1508/iStock/Getty Images Plus via Getty Images

Follow ZDNET: Add america arsenic a preferred source on Google.

ZDNET's cardinal takeaways

• Google's caller Veo 3.1 video exemplary has landed.
• It tin blend individual images into a unified video clip.
• Like its predecessor, it besides creates videos pinch audio.

Once upon a time, animators had to painstakingly activity frame-by-frame, stitching together agelong strings of still images to create nan illusion of motion. Today, they only request to upload a fewer images, and AI will do nan rest.

On Wednesday, Google DeepMind released its latest video-generating AI model, Veo 3.1, disposable now successful Flow, Vertex AI, nan Gemini API, nan Gemini App, and Vids. The institution besides released a smaller, little powerful type of nan exemplary called Veo 3.1 Fast.

Also: I utilized Google's photo-to-video AI instrumentality connected my selfie - and it made maine do nan tango

Veo 3.1 specializes successful blending disparate images into natural-looking videos, importantly reducing nan clip and resources that person historically been required for video production. Amazon besides precocious debuted an AI instrumentality which allows brands to generate short video ads from still images of products successful a matter of seconds.

Google's caller exemplary arrives little than 4 months aft nan public launch of its predecessor, Veo 3, which quickly became a deed because of its expertise to make video pinch synchronized audio. Google besides later upgraded that exemplary pinch nan expertise to generate short videos from a azygous image.

Veo 3.1 besides comes pinch that characteristic and more. According to a promotional platform from Google shared pinch ZDNET, nan exemplary "offers richer audio and enhanced realism that captures existent to life textures." It besides has a much blase "understanding of storytelling, cinematic styles, and characteristic interactions," nan institution wrote.

Video 'ingredients' 

Veo 3.1 blends aggregate images to create a single, natural-looking video, for illustration an AI blender that takes abstracted assets and combines them into a azygous ocular smoothie.

Also: Try Google's Nano Banana image generator successful Search and NotebookLM - here's how

An image of a woman's face, different of a postulation of clothing grouped together, and a 3rd of an ornate-looking room could, for example, punctual nan exemplary to create a short video clip of nan female wearing nan pictured apparel and strolling done nan room (no evidently detectable other fingers included). 

More interestingly, you tin upload images which, astatine first glance, you'd ne'er expect could beryllium brought together successful immoderate benignant of comprehensible way. This is wherever nan "creativity" (to usage a loaded term) of Veo 3.1 shines brightest. 

Want much stories astir AI? Sign up for AI Leaderboard, our play newsletter.

A demo provided by Google showing 1 image of a dressed up Christmas character down a brace of sliding doors and different of a psychedelic substance of colors -- resembling a postulation of various overgarment colors blended together -- creates a video of nan doors sliding unfastened to merchandise a flood of multicolored, Christmas ornament-sized balls, for illustration a Surrealist reimagining of nan blood-filled elevator successful The Shining.

First and past frame

Veo 3.1 besides allows users to upload conscionable 2 images -- nan first and past successful a series -- and nan exemplary will automatically capable successful nan intermediary blank spot pinch video.

Also: You tin trial Microsoft's caller in-house AI image generator exemplary now - here's how

In 1 demo video, for example, Google shows an image of an old, rustic barn, pinch debased sunlight pouring done nan entryway, and different of a cowboy astride a horse, which appears to beryllium casually trotting done gangly grass. Veo 3.1 combines these 2 images by panning nan camera done nan barn's doorway until each we spot is nan (now really moving) cowboy.

The first and past image characteristic is disposable now connected Flow, Vertex AI, and nan Gemini API, but not nan Gemini App.

Caveats

In that demo video and successful others provided by Google, some nan first and past images person akin lighting and creator aesthetics. Uploading 2 images that are wholly chopped from and unrelated to 1 different -- a achromatic and achromatic image of a Ferrari paired pinch a colour pencil sketch of an orangish tree, opportunity -- will output little predictable results.

Scene extension

Veo 3.1 besides comes pinch a caller segment hold feature, done which users tin easy lengthen their AI-generated video clips, on pinch different capacity that allows them to adhd aliases region ocular elements to and from existing videos.