Google Sneakily Launches AI App That Transcribes Everything, Even Offline: Here’s What It Really Can Do

Google has quietly released a new AI-based dictation app designed to work fully offline. The name is Google AI Edge Eloquent and it introduces a different approach compared to traditional solutions: it not only transcribes but also directly reworks the text to make it readable and ready for use.

The functioning revolves around voice recognition models based on Gemini that can be downloaded directly onto the device. Once installed, they allow the app to be used without an internet connection, with clear advantages in terms of privacy and speed. Users can also activate a cloud mode that utilizes Gemini models to further refine the text, while local processing remains always available.

Google AI Edge Eloquent does not simply report word for word what is said. It automatically eliminates fillers like “uhm” or “ah” and corrects sentences to produce smoother content that is ready to be used. At the end of the dictation, the app also allows users to rework the text in different styles to make it more concise, longer, or more formal depending on needs.

Among the most interesting features is the ability to import names, terms, and keywords from Gmail, as well as manually add custom words to enhance accuracy. The app also includes a complete history of transcription sessions with integrated search, alongside data such as dictation speed and the total number of words recorded.

Currently, Google AI Edge Eloquent is only available on iOS, although there were initially references to an Android version. Recent information suggests that Android support has been temporarily removed from the description, while a dedicated iOS keyboard is also forthcoming. Previously, Google had described a full integration on Android, allowing the app to be used as a default keyboard and to start dictation from any text field.

AI-based transcription apps are gaining more and more traction, thanks to advancements in speech-to-text models. With this experimental project, Google is testing an approach that could also come to Android in a broader form, with advanced functionalities integrated directly into the operating system.