Google quietly releases an offline AI dictation app for voice typing

Google has quietly introduced a new offline-first dictation app, "Google AI Edge Eloquent," on iOS, positioning it as a competitor to tools such as Wispr Flow, SuperWhisper, and Willow.

The app is available for free download, and once users install its Gemma-based automatic speech recognition (ASR) models, they can begin dictating directly on their devices. In the app, users see a live transcription of their speech. When dictation is paused, the system automatically removes filler words such as "um" and "ah," refining the output into cleaner, more polished text.

Users are also given options to transform the generated content, with tools labelled "Key points," "Formal," "Short," and "Long" available beneath the transcript to adjust the text's style and structure.

A key feature of the app is its ability to operate fully offline. Users can turn off cloud mode to ensure all processing happens locally on the device. When cloud mode is enabled, however, the app leverages Google's Gemini model to further enhance the text output.

The app also allows for personalisation. It can import specific keywords, names, and jargon from a user's Gmail account with permission, and users can manually add custom words to improve transcription accuracy.

Additionally, Google AI Edge Eloquent maintains a history of transcription sessions. Users can browse and search through past entries, view metrics such as words dictated in the last session, track their speaking speed in words per minute, and see the total number of words spoken.

According to its App Store description, the app is designed to go beyond basic dictation tools. It aims to convert natural speech into professional-quality text by intelligently removing hesitations, filler words, and mid-sentence corrections, delivering output that reflects the speaker's intended meaning rather than a literal transcription.

Although currently limited to iOS, the App Store listing اشاره suggests that an Android version is in development. The description highlights future Android-specific capabilities, including the ability to set the app as the default keyboard for system-wide dictation across text fields. It also mentions a floating button feature, similar to the one Wispr Flow offers on Android, that lets users access transcription quickly from anywhere on their device.

AI-driven transcription tools have been gaining traction as speech recognition technology continues to improve. With this experimental release, Google is stepping further into the space. If the app proves successful, it could pave the way for more advanced voice typing and transcription features across the broader Android ecosystem.