Atlas: Private On-Device AI

वर्ज़न 1.4.6

### What’s New **New speaker recognition model** Choose how Transcribe identifies speakers. You can now try NVIDIA’s Nemotron 3 speaker model alongside the existing model. It runs entirely on your device, downloads once on first use, and supports up to eight speakers. The existing model remains the default while we continue testing Nemotron 3 for Dutch. **One Library for everything** Search and Documents are now together in one Library. Search right from the top, revisit recent searches, and use filters to quickly find notes, email, calendar events, and more. Results are grouped by people, topics, titles, and matching passages, with search terms highlighted for easy scanning. People and Topics also have dedicated shelves on iPhone. **Settings, redesigned** Settings now gives you a clear overview of what stays on your device and what uses the cloud. See the status of your answers, Library, Spotlight, downloads, and more at a glance. A new settings search makes it easy to find anything using everyday language, such as “dark mode” or “API key.” Models & Behavior is now split into Model and Personalization, while Notifications and Storage have their own sections. **A new Briefing, Timeline, and Transcribe** Briefing now gives you a clearer view of your day, from what’s coming up and when it starts to a timeline of your meetings, people, and notes. Timeline opens with a month view showing how much happened each day, then takes you into a journal-style view of your activity. Transcribe has been redesigned with a simpler recording experience, recent transcripts, a live timer, audio activity, color-coded speakers, and automatic scrolling as you speak.