Releases: sebsto/wispr
Release list
v1.10.2
v1.10.1
v1.10.0
What's New
Meeting Transcription (experimental) — You can now transcribe meetings directly from the menu bar. Start a session and Wispr captures your microphone (and system audio when Screen Recording is granted), separating "You" from "Others" automatically. A floating transcript window shows results in real time, and you can copy or export the full transcript when you're done.
This feature is still experimental — if you run into issues, please open a GitHub issue and let us know what happened. Feedback welcome!
Built by @gbrunoo — thanks for the contribution and the patience through multiple review rounds!
What's Changed
New Contributors
Full Changelog: v1.9.2...v1.10.0
Contributors
Assets 3
v1.9.2
When you press the hotkey and release too quickly — before any speech is captured — Wispr shows a "no speech detected" notification. Previously, that notification blocked you from starting a new dictation for 5 seconds. Now, pressing the hotkey again immediately dismisses the notification and starts a new recording.
What's Changed
- Hotkey dismisses error overlay — pressing the dictation hotkey while an error notification is showing now clears it and starts recording immediately, in both push-to-talk and hands-free modes. (#52)
Full Changelog: v1.9.1...v1.9.2
v1.9.1
Fix for a bug where [TEXT_START] / [TEXT_END] delimiter tags used internally by AI text correction would sometimes appear in the corrected text, forcing users to manually remove them before sending.
Bug Fixes
- Delimiter token leak in AI text correction — The prompt sent to the Apple Intelligence model referenced the tags ambiguously (
[TEXT ...]), which could cause the model to echo them back in its response. The prompt now explicitly names[TEXT START]and[TEXT END], eliminating the issue. (#53)
Tests
- Added a regression test that calls
TextCorrectionService.correctText()against the on-device AI model with multiple inputs (English and French) to detect any delimiter token leak. (#55)
What's Changed
Full Changelog: v1.9.0...v1.9.1
Contributors
Assets 3
v1.9.0
This release adds on-device AI text correction to Wispr. After your speech is transcribed, the app can now optionally clean it up, fixing grammar, removing filler words and hesitations, and making it read more like something you'd type rather than something you'd say. Everything runs locally on your Mac through Apple Intelligence, so nothing leaves your machine.
You can turn it on in Settings. If your Mac doesn't support Apple Intelligence, the option will be grayed out with an explanation of why.
What's Changed
Full Changelog: v1.8.3...v1.9.0
Contributors
Assets 3
v1.8.3
The app no longer shows a Dock icon or application menu. Previously, some users — especially those who installed via Homebrew — would see a Dock icon and an app menu with a "Settings…" item that didn't work. Wispr is a menu bar app and all settings are accessible from the menu bar icon.
Thanks to @Tryanium and @cremich for reporting this.
Full Changelog: v1.8.1...v1.8.3
Contributors
Assets 3
v1.8.2
v1.8.1
Fixed a crash that could occur on macOS 26.3 when clicking "Install Command Line Tool…" from the menu bar. The app would quit immediately without showing the install dialog.
If you ran into this, updating to 1.8.1 should resolve it. You can also install the CLI manually in the meantime with:
sudo ln -sf "/Applications/Wispr.app/Contents/Resources/bin/wispr-cli" /usr/local/bin/wispr
Thanks to @charlesrapp for the detailed bug report and crash analysis.
What's Changed
- Fix CLI --version and archive export by @sebsto in #44
- Allow user-created aggregate audio devices in device list by @sebsto in #45
Full Changelog: v1.7.0...v1.8.1
Contributors
Assets 3
v1.8.0
Aggregate Audio Device Support
You can now select user-created aggregate devices from Audio MIDI Setup as your input source. Previously, all aggregate devices were hidden from the device picker. Internal system aggregates created by AVAudioEngine remain filtered out.
──
What's Changed
Full Changelog: v1.7.0...v1.8.0