Overview: Google has subtly introduced Google AI Edge Eloquent, a new iOS app. This free, offline-first voice dictation tool transcribes speech in real-time, eliminates filler words, and converts raw speech into polished text without an internet connection. Using Gemma-based on-device ASR models, it offers a cloud mode with Gemini for text refinement, comes without subscription fees or usage limits, and features a personal vocabulary dictionary that can import frequently used words from a user’s recent Gmail history. The app appeared on the App Store on April 6, 2026, without any formal announcement. An Android version is mentioned in the App Store listing but is not yet on Google Play.
Google launched a dictation app on Sunday without informing the public. The Google AI Edge Eloquent appeared in the iOS App Store on April 6, 2026, without fanfare, press releases, or events, resulting in a “quiet” launch. The free app doesn’t require a subscription or impose usage caps. Utilizing Gemma-based ASR models for on-device speech recognition, it keeps recordings on the phone. This feature set, considering the cost of $85 to $180 per year for popular premium dictation apps on iPhone, makes this a noteworthy release.
The Functionality
Upon opening Eloquent, users see a dictation interface with a live waveform. The app transcribes speech in real-time. It processes raw speech automatically upon pausing or stopping: filler words like “um” and “ah” are removed, and the text is refined into readable prose. The cleaned transcript is automatically copied to the clipboard for easy pasting elsewhere.
A toggle in the top-right allows switching between two processing modes. In complete offline mode, audio stays on the device and is handled by the Gemma-based ASR model locally, with no data sent to a server. In cloud mode, speech recognition starts on-device, and Gemini models perform text cleanup in the cloud. This distinction is crucial for privacy-sensitive contexts: users in regulated industries or those wary of uploading voice data to a server have a reliable local option. The increasing demand for AI tools that process data locally in 2026 rather than sending it to third-party servers has become a key factor in enterprise software selection, which Eloquent addresses in the initial user-visible toggle.
Features Offered
Besides core transcription, Eloquent features four text transformation tools: “Key points” extracts main ideas as a bulleted list; “Formal” rewrites transcripts in a professional tone; “Short” condenses text; and “Long” expands it. A history tab retains all past transcriptions, which can be individually deleted. Usage statistics track cumulative word counts and words per minute, catering to productivity-focused users interested in measuring their dictation output.
A notable secondary feature is a personal context dictionary. Users can manually add names or technical terms to improve domain-specific vocabulary transcription accuracy. Optionally, users who sign in with a Google Account can permit Eloquent to import frequently used words from their recent Gmail messages, creating a vocabulary profile without deliberate setup. This is the sole instance where Google’s broader data ecosystem appears, briefly and optionally, in an otherwise standalone local tool.
Initial iOS Release and Its Implications
Google releasing on iOS before Android is unusual. Google typically showcases new capabilities on its platform, Android, using Gemini Nano and AI Edge SDKs that run directly on Pixel and compatible hardware. Launching Eloquent on iOS first, without a corresponding Android launch, suggests the app may be an experiment in market positioning rather than a flagship rollout, or that the iOS configuration of the underlying Gemma ASR models was ready before the Android version. The App Store listing mentions an Android version, indicating parity is likely forthcoming. The sequence indicates Google launched a significant competitive product on Apple’s platform before