Google discreetly launches a free offline AI dictation app for iPhone | TNW

      In summary: Google has discreetly launched an iOS application known as Google AI Edge Eloquent, a free, offline-first voice dictation tool capable of transcribing speech in real time, automatically eliminating filler words, and converting raw dictation into refined text without needing an internet connection. The app utilizes Gemma-based on-device ASR models, features an optional cloud mode with Gemini for text refinement, has no subscription fees or usage limitations, and includes a personal vocabulary dictionary that can pull frequently used terms directly from a user’s recent Gmail history. It became available in the App Store on April 6, 2026, without any press release or announcement. There is mention of an Android version in the App Store listing, but it has yet to be released on Google Play.

      Last Sunday, Google launched a dictation app without any prior notification. Google AI Edge Eloquent appeared in the iOS App Store on April 6, 2026, without an announcement, blog post, or press event, prompting the descriptor “quietly” to be included in subsequent headlines. The app is free, does not require a subscription, and has no usage limitations. It performs on-device speech recognition using Gemma-based ASR models, ensuring recordings remain on the phone. Considering that the most popular premium dictation apps for iPhone charge between $85 and $180 annually, this combination of features is noteworthy.

      How it operates

      Upon opening Eloquent, users are greeted with a dictation interface displaying a live waveform. As you speak, the app transcribes in real-time. When you pause or stop, it automatically processes the spoken input: filler words such as “um” and “ah” are removed, and the resulting text is refined into coherent prose. The formatted transcript is automatically copied to the clipboard for easy pasting elsewhere.

      There is a toggle in the top-right corner to switch between two processing modes. In offline mode, all audio is processed locally on the device using the Gemma-based ASR model, ensuring that nothing is sent to a server. In cloud mode, speech recognition starts on the device, but text cleanup is handled by Gemini models in the cloud. This distinction is significant for privacy-sensitive situations: users in regulated industries or those wary of uploading voice data have a reliable fully local option. The increasing demand for AI tools that preserve data privacy by operating locally rather than relying on third-party servers has become crucial in enterprise and professional software procurement, and Eloquent caters to it right from the first toggle presented to the user.

      The feature set

      In addition to the primary transcription features, Eloquent offers four text transformation tools: “Key points” summarizes the main ideas into a bulleted list; “Formal” alters the transcript to sound more professional; “Short” condenses it; and “Long” elaborates on it. A history tab keeps track of all previous transcriptions, each of which can be deleted individually. The app also includes usage statistics that monitor cumulative word counts and words per minute, aimed at users focused on productivity who want to track their dictation activity.

      A noteworthy secondary feature is the personalized context dictionary, where users can add names or specific terminology to enhance transcription accuracy for particular vocabulary. Users who log in with a Google Account can also allow Eloquent to import frequently used words from their most recently sent Gmail messages, creating a vocabulary profile without needing any setup. This is the only instance in the app where Google’s broader data ecosystem is visible, albeit briefly and optionally, within a predominantly self-contained local tool.

      iOS launch and its implications

      The decision to release on iOS before Android is atypical for Google. Generally, Android is Google’s platform where it first showcases new capabilities via Gemini Nano and the AI Edge SDK operating directly on Pixel and compatible devices. Launching Eloquent on iOS first, without an accompanying Android version, indicates that this app might be more of a market positioning trial rather than a flagship product launch or that the iOS version of the Gemma ASR models was ready ahead of the Android version. The App Store listing mentions an Android version, suggesting it is forthcoming. However, this order signifies that Google has introduced a significant competitor on Apple’s platform before its own.

      The competitive effects

      The subscription model that has characterized the premium segment of AI productivity tools appears much less defensible with a technology giant like Google offering a free alternative. The two leading standalone dictation apps for iPhone, Wispr Flow and Willow, both charge $15 monthly and depend on cloud processing, while Wispr Flow streams audio through servers run by OpenAI and Meta. The most popular privacy-oriented alternative, SuperWhisper, costs $85 yearly and runs locally but is limited to Mac. Eloquent, being free, operates locally on iPhone without usage restrictions and does not require a paid plan for offline processing. For users who have paid for subscriptions primarily due to the lack of a credible free alternative, the competitive landscape has changed drastically