Google Debuts Offline AI Dictation Tool for iOS

Google has taken a quiet but deliberate step into the growing market for AI-powered productivity tools, rolling out a new offline-first dictation app designed to reshape how users turn speech into polished text.

The app, Google AI Edge Eloquent, launched on iOS without fanfare. It positions itself against emerging competitors like Wispr Flow, SuperWhisper, and Willow—a signal that voice-driven workflows are no longer niche, but rapidly becoming mainstream.

At its core, the app reflects a simple insight: people don’t speak the way they write. Professionals pause, repeat themselves, and rethink sentences mid-flow. Traditional dictation tools capture every hesitation. This one aims to interpret intent instead.

Once users download its Gemma-based speech recognition models, the app works entirely offline. Speak into your phone, and it transcribes in real time. Pause, and the software refines the output—removing filler words like “um” and “ah” while tightening phrasing into something closer to publishable prose.

That shift matters. It mirrors the difference between a rough meeting transcript and a client-ready summary—something many professionals already spend time manually editing.

Users can also reshape their text instantly using built-in options:

• “Key points” for concise summaries

• “Formal” for professional tone

• “Short” or “Long” to adjust length

The experience begins to feel less like dictation and more like collaboration with an editor.

Flexibility sits at the centre of the product. Turn off cloud mode, and everything runs locally on-device—appealing to users concerned about privacy or working in low-connectivity environments. Switch it on, and the app taps into cloud-based models from Gemini for deeper text refinement.

The app also personalises itself over time. It can import names, terminology, and frequently used phrases from Gmail, while allowing users to add custom vocabulary. For anyone who regularly dictates industry-specific language—legal, medical, technical—that feature reduces friction immediately.

Beyond transcription, the app tracks performance metrics that hint at a broader ambition. It logs session history, enables search across past dictations, and surfaces data like:

• Words spoken per minute

• Total word count

• Recently dictated phrases

That turns a simple utility into a feedback tool. How fast do you communicate your ideas? Where do you hesitate? It’s the kind of insight typically reserved for public speakers or sales professionals, now embedded in everyday writing workflows.

Google describes the product in its App Store listing as follows:“Google AI Edge

Eloquent is an advanced dictation app engineered to bridge the gap between natural speech and professional, ready-to-use text. Unlike standard dictation software that transcribes stumbles and filler words verbatim, Eloquent utilizes AI to capture your intended meaning. It automatically edits out ‘ums,’ ‘uhs,’ and mid-sentence self-corrections, outputting clean, accurate prose,”

The rollout currently limits access to iOS, but references to Android integration suggest a broader strategy. The company hints at deeper system-level functionality on Android—potentially allowing the app to replace the default keyboard and operate across any text field, much like existing voice tools.

That raises a bigger question: what happens when dictation stops being an app and becomes the primary interface?

If Google integrates this technology deeply into Android, it could redefine how users interact with their devices. Writing emails, drafting reports, even messaging colleagues could shift from typing to speaking—particularly in fast-paced environments where speed matters more than precision on the first pass.

AI transcription tools have improved rapidly in recent years, but most still require cleanup. Google’s approach targets that final step—the polish—where time is often lost.

If the experiment succeeds, expect ripple effects. Android could see smarter, built-in transcription features. Competing apps may push further into editing and summarisation. And users may begin to expect that speaking an idea once is enough to produce something ready to send.

The real test isn’t accuracy. It’s trust. Will users rely on AI to interpret not just what they say, but what they mean?

Author: George Nathan Dulnuan