Earshot 0.2.4

• Fixes slowed-down, garbled transcripts when the app's sound goes to a different output device than the system's (for example a Bluetooth headset during a call). Earshot now reads captured audio at the rate the capture device actually runs at.
• The app binary no longer includes debug symbols.

Earshot 0.2.3

• Copy works again. In 0.2.2, ⌘C in the transcript emptied the clipboard instead of copying the selection. It now copies the selected text as plain text, whether the whole transcript after ⌘A or a passage across several speakers, without the cursor of the live line. Dragging selected text into another app works too.

Install or update: download Earshot.dmg and drag Earshot to Applications, replacing the old one. First launch steps are in the README. macOS asks again for the permissions after an update, since the app is not notarized.

Requires macOS 26 or later on Apple silicon.

Earshot 0.2.2

• Select the whole transcript. ⌘A now selects every line of the transcript, even before you click in it, and a selection can run across several speakers. ⌘C copies it as plain text. Before, a selection stopped at the end of a paragraph.
• Steadier transcript window. The transcript no longer uses SwiftUI's selectable text, which could crash Earshot during a recording. New lines arriving live keep your selection in place.

Install or update: download Earshot.dmg and drag Earshot to Applications, replacing the old one. First launch steps are in the README. macOS asks again for the permissions after an update, since the app is not notarized.

Requires macOS 26 or later on Apple silicon.

Earshot 0.2.1

• Open at login. A new checkbox at the top of Settings starts Earshot when you log in. Off by default.
• macOS 27 permission name. Settings and the README now call the permission needed for Teams speaker names by its macOS 27 name, Device Control and Data Access (Accessibility on macOS 26).

Install or update: download Earshot.dmg and drag Earshot to Applications, replacing the old one. First launch steps are in the README. macOS asks again for the permissions after an update, since the app is not notarized.

Requires macOS 26 or later on Apple silicon.

Earshot 0.2.0

• Live subtitles. A small translucent window shows the last two lines said in the app you transcribe, with the speaker's name when Earshot knows it. Off by default: turn it on in Settings › Live subtitles. Drag it anywhere, change the text size from the window.
  • Fast: Apple's on-device recognition. Nothing to download, one language at a time.
  • High: Kyutai STT 1B on MLX. Handles French and English mixed, about half a second behind. 2.4 GB, downloaded only when you click Install (needs Python 3.10 to 3.12).
• One transcript file per session, so a folder indexer such as doc-indexer-macos finds each class once, next to your documents. Markdown and subtitles (.srt) now export from the Library.

Install or update: download Earshot.dmg and drag Earshot to Applications, replacing the old one. First launch steps are in the README. macOS asks again for the permissions after an update, since the app is not notarized.

Requires macOS 26 or later on Apple silicon.

Earshot 0.1.0

First public build of Earshot: live, on-device transcription of any app on your Mac.

• Pick an app in the menu bar and what it plays is transcribed as it is said, with whisper.cpp on the GPU. Nothing leaves the Mac.
• Speaker names in Microsoft Teams, read from its live captions.
• Sessions filed by calendar event, with .txt, .md, .srt and .m4a exports.
• Optional microphone, English or French interface.

Install: download Earshot.dmg, drag Earshot to Applications, then follow the first launch steps. The app is not notarized, and it needs a Whisper model.

Requires macOS 26 or later on Apple silicon.