This article is reprinted with the authorization of AIGC Studio public account, please contact the source for reprinting.

Google has announced Whisk, a new AI tool from Google Labs that uses images for a fast and fun creative process. Whisk does not generate images with long, detailed text prompts, but instead prompts with images. Simply drag in the image and start creating.

whisk summarized it as follows:

• Whisk is the latest image-generating experiment from Google LABS that focuses on quick visual ideas without having to understand the prompts in depth!
• Just add a few images for orientation reference (scene, theme, style) and Whisk will recommend some images for you to continue refining.
• Whisk is powered by Google's Gemini (a language model with visual understanding capabilities) and Imagen 3 (a model that generates images) working together.
• Turning a drawing into a stuffed animal? Making epic holiday cards? Making beautiful mood boards? Or the beginning of the story...

Build example

In the background, the Gemini model automatically writes detailed captions for your pictures. It then feeds these instructions into Google's new image-generation model, Imagen 3. This process captures the essence of the subject rather than an exact replica. Themes, scenes, and styles can be easily recombined in novel ways.

How to use it?

• Whisk the trial site: https://labs.google/fx/tools/whisk/unsupported-country
• Whisk is introduced: https://labs.google/fx/tools/whisk/faq

Instructions for use

1. You can upload 3 images, and then "Whisk" will generate an AI image that matches your criteria. If you feel that the generated AI image does not meet your expectations, you can input text as a supplement, and let "Whisk" re-generate the AI image that meets all the conditions.

1. After entering the "Whisk" page, click the "+" in the lower left corner to start generating AI images.

3. You can add 3 pictures, so that "Whisk" can generate suitable AI pictures according to your style and style.

4. The generated AI picture can be edited through text or downloaded directly.