• Fixed a potential deadlock issue when updating the configuration.

• Dictation history is now stored in a local database for fast query
• Added pagination to the history tab
• Added support for deleting individual dictation entries
• Added a confirmation dialog when clearing history

• Optimized UI (buttons, inputs, dialog overlay)
• Better prompt when recording cannot be started
• Better prompt for activation
• Better error message for license activation network error
• Recording now provides more stable real-time audio level feedback.
• Added "Don't see it?" hint in onboarding to help users find HoldSpeak in the menu bar
• Added "Manage License" button in app
• Better audio recording. Fixed latency to cancel recording

• Fixed a rare issue where switching models in settings caused the app to freeze

• You can now cancel recording by pressing the ESC button
• Fixed crashes when the model output invaded UTF-8 tokens

Enhanced the trial experience and fixed some known bugs

• System-preferred locales now appear at the top of language selection for easier access

• Introduced hands-free dictation. Pressing the hotkey less than 0.5s will lock the recorder. Short-pressing the hotkey again will stop the recording and start the transcription. Learn more here
• Added dictation history, which can be found both in the menu bar and app Settings. Learn more here

Release blog

• Improved the "Mute when recording" feature. Fixed lag when this option is toggled

• Added microphone selection
• Added option to mute system media during recording
• Added a hint at the end of onboarding to locate HoldSpeak in the menubar

• Fixed cases where commas were not followed by a space

• Added option to automatically copy output to clipboard
• Redesigned hotkey recorder UI

This release added quick settings for languages and models directly in the menubar

This release added support for binding the hotkey to the Fn key

We have also added feedback forms in the menubar. We'd love to here from you!

This release adds support for Whisper Turbo, the latest voice-to-text model by OpenAI

It is multilingual and drop in replacement to Whisper Large v3

Compare to Whisper Large v3:

• 8x faster
• 2x less parameters
• 40% VRAM efficient
• Minimal accuracy loss

To use the turbo model, simply download and enable it in the model settings

We have added a in-app feedback form

You can now send feedback to us directly from the app settings - About - Send feedback

We'd love to hear from you!

This release significantly improved performance on Apple Silicon devices

Bug Fixes:

• Fixed HoldSpeak not being triggered in some case
• Fix the model download progress not shown correctly during onboarding