What changed

The previous My Transcriber ran the UI inside a WebKit window. v0.2.0 is a real Mac app — SwiftUI for the interface, properly native UX, every macOS convention you'd expect.

The deeper change: transcription itself moved into a standalone background service that runs independently of the window. Close the window and your queued voice memos keep transcribing. The menu-bar icon shows live service status and lets you start, stop, and restart it without ever reopening the main window. A GUI hiccup can no longer wedge a transcription mid-flight — the service is its own thing.

Your data folder is untouched. Existing transcripts stay where they are. Voice Memos and Full Disk Access permissions carry over from any earlier install — you do not need to grant them again. If you were on an earlier version, the old auto-updater will replace itself with the new app the next time you launch. From v0.2 forward, updates flow through Sparkle, the de-facto macOS auto-update standard.

Faster on Intel

Intel Mac transcription used to default to Metal GPU. We benchmarked it properly — measuring the model load time and CPU% along with the transcription time itself — and found CPU with 2 threads beats GPU on Intel by about 1.9× for the default model. v0.2 ships that as the default for x86_64. Apple Silicon stays on Metal where it wins.

Both controls live in Settings → Audio transcription if you want to override the default on a specific machine.

A lot of bug fixes

Several long-standing reliability issues got tracked down and resolved this cycle — particularly around voice memo transcription edge cases, the Voice Memos capture-folder integration, and the indexing pipeline catching up after large library changes. Faster, fewer dead ends.

What's next

Not well defined yet. The immediate priority is stabilizing v0.2 through the beta — the next chapter will take shape from what holds up under real use and what feedback comes back. Bug reports from Settings → Bug report go straight into that loop.