Google, the search giant turned AI wellspring, is back with another generative AI tool called Whisk. Currently an experimental offering, it"s a new image generation tool that lets you submit images as input instead of detailed text prompts to get a remixed image in return.
You can input multiple images for various aspects, such as the subject, style, and scene. Whisk then uses Gemini to create a text prompt in the background and feeds it to Google"s image generation model Imagen 3.
For reference, the text-to-image model was launched earlier this year, and Google claims that Imagen 3 is its "highest quality image model, capable of generating images with even better detail, richer lighting and fewer distracting artifacts than our previous models."
Google explained in a blog post that its new generative AI tool "captures your subject"s essence, not an exact replica." The tool doesn"t extract all the characteristics from the input image, which could make the output image differ from your expectations. For instance, the subject of the generated image may have a different skin tone, height, or hairstyle.
Whisk lets you tweak the text prompt to make up for any considerable elements it may have lost in the generated image. "We understand these features may be crucial for your project and Whisk may miss the mark, so we let you view and edit the underlying prompts at any time," Google said.
However, the search giant added that the experimental tool isn"t a traditional image editor but more of a "creative tool" to explore new ideas and use images for a fast and fun creative process. The tool offers an option to download the generated images.
Whisk is currently rolling out to people living in the US who have signed up for the Google Labs program, which offers access to Google"s experimental generative AI features. If you"re in the US, you can access Whisk using this link.