Google recently quietly launched a new application on iOS called "Google AI Edge Eloquent". This is an "offline-first" voice dictation tool that uses the Gemma series of models to complete speech recognition locally, supports real-time transcription and automatically cleans up modal particles and redundancies, and organizes spoken speech into text that is more like formal written language.
The application is completely free, with no subscription fees and no upper limit on the number of times or duration of use. It is positioned as a productivity tool for ordinary users and a "model project" for Google to demonstrate its on-device Gemma capabilities.
In actual use, users can perform offline dictation after downloading and installing the Gemma-based speech recognition model. The interface will display speech waveforms and real-time subtitles; after pausing the recording, the application will automatically delete spoken words such as "um" and "ah" and polish the sentences. After the transcription is completed, users can convert the content into different text styles such as "key point", "formal", "short" and "detailed" with one click. If stronger polishing capabilities are needed, the optional cloud mode can be turned on, and the Gemini model will further polish the text.
Privacy and personalization are another focus of this app: in "completely offline" mode, all audio is processed locally and will not be uploaded to the server; when cloud enhancement is turned on, speech recognition is still performed locally first, and only subsequent text cleaning is completed in the cloud. Users can add common names, terms, and jargon to the built-in Dictionary feature, or choose to import high-frequency words from their recent Gmail sent messages to improve recognition rates.
Currently, Google AI Edge Eloquent is available on the Apple App Store, but only on the iOS platform. The application description mentions "seamless Android integration" many times and can be used as a system-level keyboard and floating button input layer. It is generally believed that the Android version is on the way, but Google has not officially announced a timetable. Many foreign media pointed out that this application directly competes with new generation AI dictation tools such as Wispr Flow, SuperWhisper, and Willow, and further strengthens Google's presence in the "local AI (AI Edge)" strategy.