Google may bring more features to its AI-powered Circle to Search, including a text-to-speech function, a report suggests. The visual search feature, first announced in January, allows users to circle, tap or mark an object on their screen using gestures and search for it on the web. Google announced an update to it at its developer conference on May 14, equipping the feature with math-solving capabilities. A new report now suggests that even more features could be arriving soon with three features believed to be in the works.
In a report, Android Authority, in collaboration with tipster AssembleDebug, revealed that the Circle to Search feature could soon read a selected part of the screen aloud. The discovery was allegedly made after disassembling the APK of the Google beta application v15.20.36.29. This hints at the potential text-to-speech capabilities coming to Circle to Search. Two other features – Listen and Select All – may also be introduced, the report suggests.
It has also been reported that the ‘Listen’ and ‘Select All’ features could be taken from Google Lens, where the Listen feature reads highlighted text aloud, while the ‘Select Text’ option can highlight text present in an image or visible on the camera’s viewfinder.
The report also suggests that a ‘Save’ option could be available to users after selecting an area of the search screen. Using this option would supposedly save the screenshot in the Google app’s Saved tab in the “Uploaded Images” collection instead of the device’s gallery app, which is usually Google Photos.
In theory, all of the proposed features could contribute to those recently introduced in Circle to Search. Initially only available on the Samsung Galaxy S24 series, it has made its way to other Android smartphones over the past few months.
Since its introduction, the feature has received several new updates aimed at improving its capabilities. One of them that stands out is a specialized tool that allows students to solve math and physics problems with step-by-step instructions. It uses LearnLM, a recently introduced family of learning-friendly AI Large Language Models (LLM) built on top of Google’s Gemini LLM.
