Google has listed AI Edge Eloquent on the iOS App Store, introducing a free AI dictation app built for offline speech recognition. The app runs on Google’s on-device Gemma ASR model, keeping audio processing on the handset instead of sending it online, and it arrived on iOS before Android.
Speech turns into cleaned-up text on the device
The workflow is simple. After opening the app, users speak into the microphone and see live text and waveform feedback on screen. Once recording stops, the app removes filler words such as “um” and “ah,” fixes sentence breaks, and prepares cleaner text that is automatically copied to the clipboard.
Privacy is a central part of the product. In its offline mode, audio does not leave the user’s iPhone and is not sent to any server. Google also offers a cloud mode: speech recognition still happens on the device, while only the recognized text, not the raw audio, is sent to Gemini in the cloud for additional cleanup and higher output quality.
According to the app information, the service is free to use, with no subscription fee and no usage cap.
An unusual rollout puts iOS ahead of Android
The release stands out because Google usually ships new products on Android first, especially for on-device AI features that are often tied to Pixel hardware and Gemini Nano. This time, AI Edge Eloquent appeared on iOS first, and the App Store description says an Android version is coming soon.
The source points to two readings of that decision. One is that Google may be using iOS as a testing ground to measure consumer response to on-device AI dictation without affecting its Android ecosystem. The other is that the iOS build of the Gemma ASR model may have reached a usable level earlier, leading Google to publish it without waiting for a simultaneous launch.
Gemma moves closer to everyday consumer use
From a technical angle, AI Edge Eloquent is more than a dictation utility. Gemma is Google’s open lightweight AI model family, designed to run efficiently in constrained environments such as mobile devices. AI Edge is the company’s framework for running machine learning models locally on hardware.
That makes the app a direct example of Gemma’s on-device inference moving into a real consumer setting. It also shows how an offline-first design can be paired with cloud text cleanup to balance privacy and output quality, while giving developers a practical model for how such systems can be packaged into end-user software.