Changelog

• Bug fixes and improvements

• Bug fixes and improvements

• Bug fixes and improvements

• Added Anthropic and Groq as bring-your-own-key (BYOK) AI providers for transforms.
• Fixed transforms using your own API key now returning clean text instead of raw data.
• Fixed the AI model picker dropdown being cut off in certain parts of the interface.
• Improved the API key setup area to clearly highlight when a key is missing and where it's needed.

Bug fixes and improvements.

• Added Anthropic and Groq as bring-your-own-key provider options.

• Bug fixes and improvements

• Bug fixes and improvements

• Bug fixes and improvements

• Bug fixes and improvements

• Bug fixes and improvements

• Bug fixes and improvements

• Bug fixes and improvements

• Bug fixes and improvements

• Bug fixes and improvements

• Bug fixes and improvements

• Bug fixes and improvements

• Bug fixes and improvements

• Fixed transcription quality degrading after repeated use by preventing CPU overcrowding between the transcription engine and other app components.

• Improved transcription reliability by preventing slowdowns that could occur after repeated use.

• Bug fixes and improvements

• Improved voice detection accuracy to better distinguish speech from background noise.
• Fixed an issue where very long recordings could cause transcription to slow down or fail.

• Added a Preferred Language setting to ensure transcriptions always come out in your chosen language, even when the app gets confused and outputs the wrong one.
• Fixed an issue where empty transcription results would not trigger the automatic restart, causing the app to get stuck after repeated failures.

• Added a Preferred Language setting that ensures transcriptions are always output in your chosen language, fixing cases where speech in one language was incorrectly transcribed in another.

• Fixed a delay at the start of your first recording after launching the app.
• Fixed an issue where audio could be lost during long recordings.
• Improved transcription reliability for long recordings (2 minutes and beyond).

• Improved transcription reliability so no spoken audio is ever lost, even in long recordings.
• Fixed an issue where recordings over 90 seconds would have audio cut off and discarded.
• Improved accuracy for long dictation sessions by processing speech at natural pause boundaries.

• Fixed the overlay status label to appear below the waveform and now shows 'Processing…' and 'Transforming…' during the appropriate stages.
• Fixed an issue where very long local recordings could cause slow or degraded transcription by capping audio at 90 seconds.

• Improved the dictation overlay to show status labels below the waveform, with clearer "Processing…" and "Transforming…" indicators during each step.

• Fixed an issue where transcription quality could silently degrade after extended use, causing the app to automatically refresh itself for a clean slate.
• Improved status indicator to only appear when there is a meaningful extra wait, reducing unnecessary visual noise.

• Improved the recording status display to only show an explanatory label during the transcription processing step, reducing visual clutter.

• Fixed a visual flash in the overlay that occurred between transcription and text transformation.
• Added a status label under the visualizer showing whether Koe is processing or transforming your speech.
• Improved transcription stability by ensuring the app fully restarts after updates, preventing quality degradation over time.

• Fixed an issue where transcription quality could degrade after installing updates.
• Improved app restart behavior after updates to ensure a clean, fresh start every time.

• Fixed an issue causing transcription quality to degrade after a few uses.

• Improved local transcription accuracy by using smarter silence detection that handles whispers, varying volume, and background noise.
• Improved local transcription speed by reducing the amount of audio processed — silent portions are now stripped before transcription.
• Fixed an issue where repeated transcriptions would slow down and produce garbled text over time.
• Added a quality check that detects and discards garbled transcription output instead of pasting it, showing a clear error message instead.

• Improved local transcription accuracy by better detecting and removing silence before processing speech.
• Fixed an issue where garbled or low-quality transcription results could be pasted into your app — these are now rejected with a clear error message instead.

• Improved transcription accuracy for recordings with long pauses by automatically removing silence before processing.
• Fixed an issue where transcribing long audio clips (over 10 seconds) could return empty text.
• Fixed a crash that occurred when transcribing text containing accented characters (é, à, etc.).

• Fixed a crash that occurred when transcribing text containing accented characters (such as French é, à, etc.), which also broke all subsequent transcriptions in the session.
• Fixed an issue where long audio recordings (over 10 seconds) returned empty transcriptions.

• Bug fixes and improvements

• Fixed transcription quality degrading after several recordings in a row.

• Bug fixes and improvements

• Fixed an issue where local transcription quality would degrade after several uses in a row.

• Fixed transcription quality degrading over time during a session.
• Fixed opening files in your chosen code editor from the editor picker.
• Removed terminal apps from the editor picker, which would run files instead of opening them for editing.

• Fixed an issue where opening files with an external editor could fail due to system permission restrictions.
• Removed terminal apps (such as Warp and Ghostty) from the editor picker, as they would try to run files instead of opening them for editing.

• Fixed an issue where opening files with the editor picker could fail on some systems.
• Removed terminal apps from the editor picker, so only code editors and IDEs are shown.

• Fixed the "Open hooks.json" button, which was previously blocked and did nothing.
• Improved the hooks settings layout for a cleaner, easier-to-use interface.
• Added automatic detection of installed code editors (Cursor, VS Code, Zed, Sublime Text, and more) with a dropdown to choose your preferred editor for opening hooks.json.

• Fixed layout of the hooks settings panel, with dropdowns and inputs now properly arranged.

• Added a Hooks management section in Settings to view, add, and remove hooks without editing files manually.
• Fixed a delay that caused voice-to-text to take several extra seconds to process.
• Fixed an error that appeared on every app launch related to API key setup.
• Improved error visibility in exported logs to make troubleshooting easier.

• Fixed an error that occurred on every app launch related to API key migration.
• Fixed a delay of several seconds that affected text transformation after dictation.
• Improved error visibility in exported logs to help diagnose issues.

• Fixed a 3-second delay that occurred before every transcription transform.
• Improved internal error logging to make issues easier to diagnose from exported logs.

• Fixed an issue where text transformation could hang indefinitely after dictation.
• Fixed a bug where the app could silently fail on certain operations without any indication of what went wrong.
• Improved error visibility so that issues can be more reliably diagnosed from exported logs.