Google LLC today introduced image and text generation features for Search Generative Experience, or SGE, an experimental version of its search engine powered by artificial intelligence.

SGE made its debut in May at Google’s annual I/O developer event. When a user enters a query into the service, it displays standard search results alongside a panel at the top of the interface. The panel contains a natural-language answer to the user’s question and, in some cases, also includes explanatory visuals such as diagrams.

Google has updated SGE with new features since its introduction four months ago. One of the most recent upgrades, which rolled out in August, added a tool that can automatically summarize online articles. Google also enhanced SGE’s existing code generation capability.

The update announced today adds two features that will make it possible to generate images and text, respectively, with natural-language prompts.

Activating the image generation tool requires users to enter a short description of the object they wish to draw. A user could, for example, type in the prompt “draw a smartphone on a table.” SGE will then generate up to four images that correspond to that description.

Although they share key details, the images the service generates based on a prompt differ in certain respects. In response to the prompt “draw a smartphone on the table,” SGE might generate one image that shows an iPhone and another featuring an Android handset. There can also be more subtle differences.

SGE provides a short, natural-language description of each image that it creates. This description, which is different from the user prompt that was used to create the file, contains a detailed overview of the pictured objects. Users can change this overview by replacing certain words or adding new ones and the changes will be automatically applied to the image. 

Google plans to include a watermark in each SGE-generated image to indicate that it was created by AI. For added measure, the search giant will add a metadata label designed for the same purpose.

“If you’re opted in to SGE, you may also see an option to create AI-generated images directly in Google Images, as part of this experiment,” Hema Budaraju, Google’s senior director of product management for search, wrote in a blog post. “This feature is designed to appear when you’re searching for inspiration, like ‘minimalist halloween table settings’ or ‘spooky dog house ideas.’”

The feature is rolling out alongside an AI tool that makes it possible to generate prose with short natural language prompts. Users can, for example, ask the tool to write customer support tickets or a short essay. Built-in customization options make it possible to vary details such as text length and tone. 

“From there, it’s easy to export your draft to Google Docs or Gmail when you’re ready to add that personal touch,” Budaraju wrote. “And rest assured that Google Workspace privacy protections carry over into this export functionality.”

Both the image and text generation features are initially rolling out to SGE users in the U.S. with support for English-language queries. SGE is accessible to participants in Google’s Search Labs product testing program. That program, in turn, is open to eligible users who choose to sign up through a button embedded into Chrome’s new tab page.

Image: Google