Google has quietly rolled out a new iPhone app called Google AI Edge Eloquent, a free dictation tool that runs primarily on-device and is designed to turn messy spoken input into cleaner, more polished text. The low-key launch is notable not just because it happened on iOS first, but because it gives users a no-cost, offline-first speech-to-text option at a time when privacy, latency, and subscription fatigue are shaping the voice AI market.
TL;DR
- Google AI Edge Eloquent is a new free iPhone dictation app that works offline for core speech-to-text tasks.
- The app uses Gemma-based, on-device models, removes filler words, and can polish transcripts into cleaner prose.
- It is currently iPhone-only, weighs 67 MB on the App Store, and Google’s listing says keyboard support is coming soon.
- Optional cloud features exist, but Google says core machine learning processing runs locally on the device.
What Did Google Launch?
Google’s latest AI experiment arrived without the kind of blog post, keynote mention, or product fanfare usually attached to its consumer launches. Instead, Google AI Edge Eloquent appeared on Apple’s App Store on April 6, 2026, as a free Productivity app for iPhone. The listing shows it is developed by Google, available in English, sized at 67 MB, and updated through version 1.2.2 within hours of launch, pointing to an active early rollout.
At a basic level, Eloquent is a speech-to-text app, but Google is positioning it as something more polished than standard dictation. The company says the app is meant to bridge the gap between natural speech and professional, ready-to-use text, instead of simply transcribing every pause, stumble, and correction word for word.
How Does AI Edge Eloquent Work?
According to Google’s App Store description, Eloquent is powered by Google’s Gemma architecture and uses efficient, open-weight, on-device models for speech-to-text. Google says the app can automatically remove filler words, clean up mid-sentence corrections, and output more readable prose.
The app shows live transcription as users speak and then refines the text when recording is paused. It also includes rewrite options such as Key points, Formal, Short, and Long, giving users a way to reshape rough dictation into summaries or more presentable copy.
What About Privacy And Cloud Processing?
Privacy is one of the strongest hooks in the launch. Google says all machine learning processing runs entirely locally on the iOS device and that audio, confidential conversations, and personal data do not leave the device, although the listing also notes that some advanced optional features may require the cloud.
That matches reporting that users can switch off cloud mode for local-only processing, while cloud mode uses Gemini models for text cleanup. In other words, Google is giving users both a privacy-first local option and a more enhanced cloud-assisted mode.
Why Does This Launch Matter?
The significance of Eloquent goes beyond dictation. Google has spent the past year building out its AI Edge and LiteRT stack to make Gemma and other models run efficiently on-device. In January 2026, Google said LiteRT supports the Gemma family and pointed developers toward AI Edge Gallery experiences on both Android and iOS, reinforcing its wider push to make local AI practical across devices.
Eloquent looks like one of the clearest consumer-facing examples of that strategy so far. Instead of only showing developers what on-device AI can do, Google is now packaging those capabilities into a real-world productivity app that directly competes with paid dictation tools. That could make Eloquent both a test product and a proof point for Google’s edge AI ambitions.
Topics for more insights:
Why Is The iOS-First Rollout Interesting?
The platform choice stands out too. Eloquent is only available on iOS right now, even though Google’s AI Edge efforts are often associated more closely with Android. The App Store listing also references broader integration plans, suggesting that Android support may follow later.
That makes this launch feel slightly unusual for Google. Releasing an experimental AI productivity app on Apple’s platform first could signal that the company is testing demand quietly before expanding further, or simply that this iPhone build was ready first. Either way, it is an interesting move in a voice AI market that is increasingly being shaped by on-device performance and lower-cost usage models.
Join The Discussion