Google Quietly Enters Offline AI Dictation With New Eloquent App

Google has quietly released a new dictation app that pushes generative AI toward something more practical than chatbots and flashy demos: fast, offline speech-to-text on a phone.

The app, called Google AI Edge Eloquent, appeared on iOS on April 7, 2026, and is designed to transcribe speech locally after users download its Gemma-based models. It also offers optional cloud-based cleanup through Gemini when users want polished output.

A Dictation App Built for Real-World Use

Eloquent’s main pitch is simple. Users speak, the app transcribes, and the result is cleaned up into readable text. According to the App Store description and TechCrunch’s reporting, the app removes filler words such as “um” and “ah,” then offers writing styles including Key points, Formal, Short, and Long.

That puts Google in a growing category of AI-powered dictation tools that aim to do more than convert voice into text. The emphasis here is not just accuracy, but turning rough spoken notes into something ready to send, save, or share.

Why Offline Matters

The offline-first approach is the more notable part of the launch. By running speech recognition locally after the model download, Eloquent can work without a live internet connection. Google also says users can switch cloud mode off entirely for local-only processing.

That matters for people who want faster response times, fewer connectivity worries, or more control over where their audio is processed. It also reflects a broader shift in generative AI product design, where companies are increasingly trying to make AI useful on-device rather than only through remote servers.

What Google Is Testing Next

The app is currently available on iOS, but its App Store description also references an Android version and “seamless Android integration.” TechCrunch reported that Google later updated the listing to remove references to an Android app while adding that an iOS keyboard is coming soon.

For now, that makes Eloquent feel like an early test case rather than a finished platform rollout. Still, it gives a clear signal about where Google sees consumer AI heading: less about novelty, more about quietly useful tools that save time in everyday workflows.

Key Takeaways

• Google released an offline-first dictation app called Google AI Edge Eloquent on April 7, 2026.
• The app uses Gemma-based speech recognition and can optionally use Gemini in the cloud for text cleanup.
• It removes filler words and offers output styles such as Formal, Short, Long, and Key points.
• The launch suggests Google is testing practical, device-first generative AI features for everyday use.

What to Watch

The main question is whether Google expands Eloquent beyond this early iOS release and into a broader consumer product. If the company keeps the offline transcription quality high, the app could become a useful example of how generative AI moves from abstract assistant to everyday utility.

Also worth watching: whether competitors respond with stronger offline dictation features of their own, and whether Google folds Eloquent’s capabilities into other products, including Android and the keyboard experience it appears to be testing next.