Google quietly launched an AI dictation app that works offline

Google has stealthily entered the competitive AI dictation space with the launch of 'Google AI Edge Eloquent,' a free iOS application. The app's core feature is its offline-first capability, powered by local Gemma-based automatic speech recognition (ASR) models that users download to their device. This ensures privacy and functionality without an internet connection. Once a dictation session is paused, the app's AI automatically edits out filler words like 'um' and 'ah,' delivering polished, ready-to-use text. Users can further refine transcripts with one-tap transformations to generate 'Key points,' 'Formal,' 'Short,' or 'Long' versions of their speech.

The app includes practical features for professional use, such as the ability to import specific keywords and jargon from a connected Gmail account and add custom words to a personal dictionary. It provides detailed session analytics, tracking words-per-minute and total words spoken. While currently iOS-only, the App Store description hints at a future Android version promising 'seamless Android integration,' potentially acting as a default keyboard and using a floating button for system-wide access. An optional cloud mode leverages Google's more powerful Gemini models for enhanced text cleanup when privacy is less of a concern. This experimental release positions Google to test advanced transcription features that could eventually be integrated across its ecosystem, directly competing with apps like Wispr Flow and SuperWhisper.