- Updated: April 6, 2026
- 2 min read
Google AI Dictation App Launches Offline‑First iOS Voice Transcription Powered by Gemma Models
**Google quietly released an offline‑first AI dictation app for iOS**, dubbed **Google AI Edge Eloquent**. The new service offers live transcription, automatic filler‑word removal, and a customizable vocabulary while letting users choose between cloud‑based processing and on‑device inference. Powered by the latest **Gemma AI models**, the app promises high‑accuracy speech‑to‑text without an internet connection, a first for Google’s consumer AI suite on iOS.
### Offline‑First Architecture
Eloquent stores a lightweight version of the Gemma model locally, enabling real‑time transcription even when the device is offline. When a network is available, the app can switch to a cloud‑enhanced mode for even richer language understanding. This dual‑mode approach mirrors Google’s broader Edge AI strategy and hints at future Android support.
### Competitive Landscape
The launch positions Google against emerging competitors such as **Wispr Flow**, which also markets on‑device transcription but relies on older model architectures. By leveraging Gemma, Google claims lower latency and higher accuracy, especially for niche vocabularies in professional settings.
### Integration with UBOS Ecosystem
For readers interested in building AI‑driven voice solutions, explore our related guides:
– [AI Dictation](/ai-dictation)
– [Google AI](/google-ai)
– [Voice Assistants](/voice-assistants)
These internal resources dive deeper into implementation details, best practices, and how to integrate Google’s Edge models with existing workflows.
### What’s Next?
Google hinted at expanding the offline‑first capability to Android later this year and hinted at tighter integration with its broader AI ecosystem, including Gemini‑based assistants.
For the full story, see the original report on [TechCrunch](https://techcrunch.com/2026/04/06/google-quietly-releases-an-offline-first-ai-dictation-app-on-ios/).
*Stay tuned to UBOS for the latest updates on AI transcription, voice assistants, and edge computing.*