Google launches Whisk, an AI that remixes photos and creates new images

Google this Monday (16) rolled out updates to the company’s artificial intelligence (AI) platforms that deal with photos and videos. Additionally, the company has come up with a new experiment to help you create fun illustrations that use as little text as possible.

It is called service Whiskeya website you can make Create new images based on photos saved on your PC or mobile phone gallery — What Google calls “remixing” content.

The company’s goal is to provide practically automatic renderings, allowing you to use text descriptions only when necessary or get more accurate results.

The whisk works based on three steps. First, you choose which one is the main item of the image, such as a person or object. Immediately, It’s time to choose the scriptsuch as a landscape or a monochrome background.

The final stage includes: describe the creation style – among many other possibilities, such as surrealist painting, science fiction aesthetics or animated films.

Whisk is currently in experimental phase and Alpha release; This means the results are not that accurate yet and error messages are still common. It’s likely that the tool won’t be available for long, as it’s considered a trial by Google and could be incorporated into existing services like Gemini in the future.

Just click here to access Whisk. The service is not currently available in Brazil, but it is possible to access the service through a virtual private network (VPN).

More news from Google in AI

In addition to Whisk, Google also detailed updates to two of the company’s AI services that deal with visual materials. They have been developed to provide more complex and accurate results, as well as the understanding of more precise commands full of references.

One of the developments Veo2The next generation of the company’s generative video language model. This is a competitor to services like Sora”understand real world physics And Human nuances of expression and movement” To create a short scene from scratch. You can even request specific technical specifications, such as framing and lens type, using professional cinematography language.

This generative AI model Works experimentally in VideoFXIt is a feature that is still in the testing phase and offered to a limited number of users.

Finally, Google has also improved the rendering of static images from the language model Picture 3. that now renders figures in more artistic styles, with richer detail and less hallucination. Also works experimentally from VideoFX.