🗂️ Generate OnePagers in the Background (opt-in)#
Prefer not to wait on a blocking dialog every time you generate a summary? Flip on Generate OnePagers in the Background in Session settings and generation moves to a single background queue, so you can keep taking notes, switch sessions, or line up another summary while one is cooking. Progress shows in the Background panel instead of locking the app. It's off by default, so head to the setting to turn it on. Either way, the moment you trigger a generation you get instant visual feedback that the click registered.
🎛️ Choose How Each AI Feature Runs#
Settings now let you pick, per feature, how the AI behaves: run it on Gemini, hand it to a command-line AI agent (the CLI engine defaults to Claude Code, which Azynote auto-detects and runs for OnePagers and Command Panel transforms), or turn it Off entirely. The AI settings were rebuilt around this, so it's clear at a glance what's powering your summaries, transcript cleanup, and chat, and features you haven't set up simply stay out of your way.
📣 See What's New Without Leaving the App#
There's a new What's New panel (tap the megaphone) that surfaces the latest features and onboarding videos right inside Azynote. It's fed from the Azynote blog, so it stays current on its own, no update required to catch the newest tips.
⏱️ A Nudge When You Forget to Stop#
Ever wrapped a meeting but left Azynote transcribing for another hour? When the room goes quiet for a while, a gentle floating reminder now asks whether you meant to stop the session. No auto-stop, no nagging notification, just a quiet check-in you can dismiss.
✂️ Cleaner Transcript Editing#
Two small but handy additions: a Copy transcript button right in the editor toolbar, and a delete-sentence action (with a trash affordance in the inspector) so you can prune a stray line without hunting through the text.
🐛 Notable Fixes#
4 fixes
- Long transcriptions run cooler: a self-tuning throttle cuts sustained CPU and GPU load during heavy sessions.
- The capture heartbeat is now inaudible: the keep-alive tone dropped below hearing range and fades in gently.
- Diarization is steadier across sessions: in-process speech models are released between runs, so back-to-back meetings stay stable.
- Command Panel now shows a clear re-auth prompt when Vertex credentials expire, and its spinner actually renders while a transform runs.
Happy meetings! 🎙️