AI was the keyword at this year’s Google I/O developer conference, with Google showcasing how generative language models are changing some of its key products. It looks like the company is gearing up to add that same magic to Gboard, pretty much the best Android keyboard app around. Hidden flags within the app make it possible to activate a range of generative machine learning features like advanced spell check and emoji generation.
As spotted by @AssembleDebug on Twitter (or X, if we must) and 9to5Google, there are several features that Google is working on in the latest Gboard beta, 13.3. The most interesting one is probably Proofread with Gboard. When activated, a popup appears that explains, “When you use Proofread, the text you write, your feedback, and your interaction with the feature will be sent to and stored on Google servers for 60 days to provide you with proofreading suggestions and help improve this feature.”
In practice, a chip will appear at the top of the keyboard after you finish a sentence, offering a “Fix it” option that’s also accessible via a toolbox shortcut. Tapping it replaces the keyboard with different suggestions, with the specific tweaks highlighted. In AssembleDebug’s example, their incorrect “What are wrong with gboard?” is corrected to either “What is wrong with gboard?” or “What’s wrong with Gboard?” The two options are accompanied by thumbs up and thumbs down buttons, and Google prominently asks for feedback (ironically incorrectly adding a question mark to the imperative phrase “Tell us how we’re doing?”). A tap on one of the suggestions replaces the original text with it.
On top of proofreading, 9to5 spotted that Gboard could get a feature similar to Help me write in Gmail, with a string saying “Select the style or tone you’d like to rewrite with.” That way, you can use generative machine learning to make your writing more formal or informal, with other options likely also available.
Other than the correction feature, you can also look forward to an “Emogen” sticker generator. It will allow you to create emoji style stickers with generative machine learning. The interface isn’t finished yet, but based on what is visible right now, it will look similar to emoji and GIF search, with a bar for prompt entry at the top.
Gboard 13.3 also prepares a few regular features that don’t leverage generative machine learning. Stylus users are in for a lot of treats, with them receiving a separate “Write in text fields” option. As the name implies, you will be able to write into any text field, with Gboard automatically turning the handwritten text into machine-readable letters. You can also scratch out text to delete it, insert text into a specific position by drawing an arrow or caret, draw vertical lines to join separate words, and circle text to select it. There are some more options, like creating a new line, but it’s not entirely clear which gestures you will need to use for this just yet.
Those of us who prefer to use voice typing will be able to hide the keyboard while dictating, leaving only the toolbar with the microphone and other options in view. The string 9to5 spotted tells us that “You can easily move the toolbar anywhere when voice typing.”
Finally, Google is also adding support for predictive back animations with this release, which will make navigating Gboard’s settings pages feel more natural on Android 14.
As with any features spotted in development, it’s not clear when Google will launch them. We wouldn’t be surprised if the company timed them to coincide with the Android 14 launch, slated for later this month, or for one of the famous feature drops that Google releases every three months following a big Android release.