Google quietly launched an AI dictation app that works offline

Google has quietly introduced a new offline-focused dictation app named Google AI Edge Eloquent for iOS, stepping into competition with tools like Wispr Flow, SuperWhisper, and Willow.

The app is available for free and begins working once users download its Gemma-powered automatic speech recognition (ASR) models. It enables real-time voice-to-text conversion, displaying live transcripts as you speak. When paused, the app intelligently refines the output by removing filler words such as “um” and “ah,” delivering cleaner, more polished text.

Users can further customize the output using built-in options like “Key Points,” “Formal,” “Short,” and “Long,” allowing flexible formatting based on their needs.

A key highlight is its offline capability—users can disable cloud processing and rely entirely on on-device functionality. When cloud mode is enabled, the app leverages Gemini-based models to enhance text refinement. Additionally, it can pull relevant keywords, names, and terminology from Gmail (with permission), while also allowing users to add custom vocabulary.

The app keeps a record of past transcription sessions, making it easy to search and review previous entries. It also provides useful insights such as words-per-minute speed, total word count, and recently dictated content.

According to its App Store listing, the app is designed to convert natural speech into professional-quality text. Unlike traditional dictation tools that capture every hesitation, it focuses on understanding intent—removing filler words and mid-sentence corrections to produce clear and accurate output.

Currently limited to iOS, the description hints at a future Android release. It also mentions potential features like system-wide integration as a default keyboard and a floating shortcut for quick access, similar to existing Android transcription tools.

With speech-to-text technology rapidly improving, AI-driven dictation apps are becoming increasingly popular. Through this experimental launch, Google is entering the space, and if successful, it could lead to more advanced transcription capabilities across Android devices in the future.

Voice of Osiz

At Osiz, we see Google’s launch of an offline-first dictation app as a strong signal toward the future of edge AI and real-time intelligence. By enabling on-device speech processing, this innovation enhances privacy, speed, and user control—key pillars for next-gen AI applications. The integration of advanced ASR models with intelligent text refinement showcases how AI is evolving from simple transcription to context-aware communication tools.

As AI adoption accelerates, businesses must focus on building low-latency, secure, and scalable AI solutions that operate seamlessly across devices. This trend also highlights the growing importance of AI-driven productivity tools in enterprise ecosystems. At Osiz, we empower organizations to harness such innovations through customized AI development strategies. The future lies in intelligent, offline-capable AI systems that deliver real-time value.

Source: TechCrunch.com