Google brings fast, private dictation to iOS with on-device Gemma models
Google has quietly launched an offline-first AI dictation app for iOS that runs its Gemma family of models on-device. By prioritizing local inference, the app enables real-time speech-to-text without needing a constant internet connection — a win for users who value privacy, reliability, and low latency.
The app's on-device approach means transcriptions happen quickly and stay on the user's device unless they choose otherwise, reducing exposure of sensitive audio to cloud servers. That makes it especially useful for journalists recording interviews, field workers in remote areas, and anyone who needs accurate dictation without relying on strong network coverage.
Key benefits include:
- Offline transcription powered by Gemma models for faster response times and reduced network dependency.
- Enhanced privacy since audio and text can remain local to the device by default.
- Broader accessibility: improved tools for people with hearing, speech, or motor challenges and for multilingual users in noisy or low-connectivity settings.
Beyond individual benefits, the release signals more competition in mobile dictation, nudging rival apps to improve privacy, speed, and offline capabilities. Overall, Google's offline-first dictation app is a practical, user-centered step that makes powerful AI-driven speech tools more usable and trustworthy across varied real-world situations.