v0.6.0-beta6

Released 2026-06-13 · Download .dmg (34 MB)

A polish release. Five fixes that, taken together, are the difference between “looks broken when I plug in headphones” and “feels like a finished app.” None are individually critical — if beta5 has been working for you, beta6 is just nicer.

AirPods plugged in after launch now record cleanly

Symptom: launch Minutiva Private with only the built-in mic available, then connect AirPods (or any USB / wired headset), then start a recording. Beta5 would either keep recording from the now-stale built-in mic, capture silence, or surface a confusing “Bluetooth voice mode” error.

Root cause: AVAudioEngine’s input node lazily binds to whatever device is system-default the first time you access it, and on macOS the engine doesn’t transparently re-route when the default input changes later in the session. The reused engine kept its original binding. Fix: rebuild the engine at the top of every start() call so it queries the current default input fresh. Two-line change that closes a whole class of “wrong device after device swap” bugs.

The Library sidebar can no longer slide off the left edge

Symptom: open the app on a narrow window with the Other voices panel showing, and the Library sidebar would get nudged off-screen to the left — meeting titles cut to …ique 1:1 instead of Rafa:minique 1:1. Clicking the play button on the audio scrubber sometimes did nothing because the hit-test rect was outside the visible window.

Root cause: NavigationSplitView’s column-width minimums are advisory — when sidebar (320) + detail content + speaker panel (280) sum past the actual window width, macOS slides the leading column off the screen instead of widening the window or hiding a pane. Fix: hard window minWidth of 1100pt so all three panes physically fit before the user can resize smaller.

The “Other voices” panel no longer jumps in mid-render

Symptom: click a meeting, the layout would settle, then half a second later the right-hand speaker panel would pop in and push everything around as the speaker registry finished loading from disk.

Fix: the panel slot now mounts on first render based on synchronous flags (premium + meeting has embeddings). The async registry load no longer changes the layout — the panel itself shows a loading state while it waits. Side benefit: there’s now a sidebar-right toolbar button to toggle the panel on or off, with the choice persisting across launches.

Voice samples are actually audible now

The play button next to each speaker on the right panel used to play a 3-second sample. That turned out to be too short to recognize a voice — testers reported clicking the button and not being sure if anything had happened. Three changes:

• Sample length: 3 seconds → 15 seconds.
• The audio file open now runs in the background, so the button doesn’t feel dead while a long recording mmaps. A spinner shows during the load.
• Click the button again during playback to stop early. The icon flips to a stop symbol while playing.

What carried over from beta5

• installTap fix so the “Starting recording…” window doesn’t hang on audio device transitions
• Background-preloaded transcription model
• Opt-in “Keep model loaded between recordings”
• Live speaker labels during recording (Premium, live mode)
• People diagnostics + bulk re-match

Nothing changed in

• WhisperKit transcription pipeline
• VBxHMM clustering algorithm
• People model + file formats
• License code format

Auto-update

Existing beta channel testers get this within 24 hours. Or trigger now: Minutiva Private → Check for Updates…