Google Launches AI Edge Eloquent for Polished Offline Dictation on iPhone

Google has quietly launched a new productivity application named AI Edge Eloquent for the iPhone and Mac. This tool uses on-device artificial intelligence to transcribe spoken words and automatically remove verbal stumbles. Unlike traditional dictation software, it aims to produce ready-to-use prose by filtering out filler words like "um" and "uh" in real time.

The application is powered by Google's Gemma architecture, a set of open-weight models optimized for mobile hardware. By running the machine learning pipeline locally, the app provides responsive speech-to-text without requiring an active internet connection. This offline-first design ensures that audio recordings and confidential data never leave the user's device.

Users can customize the output by selecting different tones and lengths, such as transforming a long recording into a concise set of key points. The app also features a personal context dictionary that learns from a user's unique vocabulary and Gmail history. This integration helps the AI correctly identify industry-specific jargon and personal contact names.

Currently, Eloquent is available for free on the Apple App Store with no usage limits or subscription fees. It requires iOS 16.0 or later and is compatible with M1-based Macs and visionOS. Google has indicated that a dedicated keyboard integration is coming soon to allow the tool to work across other iOS applications. While the core features operate locally, an optional cloud-enhanced mode is available for more complex processing using Gemini models. The app is currently not available in certain regions, including the UK and the European Economic Area, as it awaits regulatory approval. This release highlights Google's commitment to delivering advanced AI capabilities directly to consumer hardware.