Skip to content

Voice-to-text dictation app with local (Nvidia Parakeet/Whisper) and cloud models (BYOK). Privacy-first and available cross-platform.

docs: update changelog for v1.6.3 with voice recognition and Hyprland changes

Gabriel Stein634e7cd9765e
diff --git a/CHANGELOG.md b/CHANGELOG.md
index 4ab03abde..50b0cea49 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -18,6 +18,17 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
- **better-sqlite3 12**: Upgraded from v11 to v12 for Electron 39 V8 compatibility
- **Localized in all 10 languages**: All permission copy changes translated across en, pt, de, es, fr, it, ru, ja, zh-CN, zh-TW
+### Added
+
+- **Hyprland Wayland Support**: Native global shortcut support for Hyprland using `hyprctl` keybindings + D-Bus, matching the existing GNOME Wayland approach (#416)
+
+### Fixed
+
+- **Soft Voice Recognition**: Enabled Auto Gain Control (AGC) for dictation microphone input to automatically boost quiet speech — previously disabled, now matches meeting mode behavior
+- **OpenAI Realtime VAD Sensitivity**: Lowered voice activity detection threshold from 0.5 to 0.3 (both client and API) so soft-spoken audio is no longer missed
+- **Speech Onset Clipping**: Increased VAD prefix padding from 300ms to 500ms to capture the quiet beginning of soft speech that was previously cut off
+- **Wayland Clipboard Paste**: Fixed `wl-copy` failing silently due to 1ms `spawnSync` timeout killing the fork before it completed — increased to 50ms (#416)
+
## [1.6.2] - 2026-03-11
### Added