Regardless of all the text translation tools out there, Google Translate is clearly the most well-recognized (and well-utilized) one. Not only does it boast residing in prime real estate on the Google app, it’s also deeply integrated into Google Lens and Circle to Search (CtS) for quick real-world and on-screen translations.

The tool, up until recently, has thrived because of its central placement, but Google wants to change that with merit-based performance. The tech giant is working on a Duolingo-like Practice feature to gamify learning new languages, paired with a new Conversation mode, and a broader AI-based refresh for more accurate translations. Some of those updates are now beginning to take shape.

First spotted and reported on by credible Android analyst AssembleDebug in an Android Authority report, the app’s latest version 9.15.114 build offers new code, which, when manually enabled, paints a clearer picture of what some of the upcoming Google Translate features and UI tweaks would look like.

For starters, the app is moving some controls around to make way for new ones. Practice, for reference, will occupy the bottom right position, with the camera icon moving to the center, and the mic icon now appearing within the text input box. The icon will be accompanied by buttons for Paste and handwritten translations in the same text input box, as seen in the second screenshot below.

Speed or accuracy?

You’ll get to decide

Current (left), potentially upcoming (right)

More importantly though, eagle-eyed users would have spotted another key feature addition in the image above: the inclusion of a dedicated model selector.

Available via the app’s home screen and the results page, the app will soon let you swap between Fast and Advanced modes for all your translation needs. The former is optimized for “speed and efficiency,” while the latter “specializes in accuracy using Gemini.”

Fast would likely be useful for simple everyday needs, like translating a menu, or a sign when you’re in a foreign country. Advanced, on the other hand, will be more nuanced, using Gemini to understand context before translating content.

In its current implementation, the Advanced translation model is limited to English-Spanish and English-French language pairs. If and when finally available, the model could support more language pairs.