v1.7.0
- Faster Dictation by Default: Semantic finalization is now off by default, so recognized text is written immediately. Users can still enable formatting or translation modes when needed.
- Lower Dictation Latency: Translation and finalization can pre-process while recognition is still running, and Enter can skip post-processing after release to write the recognized text directly.
- Compact Dictation Indicator: The dictation indicator is smaller with less vertical padding while still supporting two-line previews for longer text.
- Chinese Recognition Default: New installs now default recognition language to Chinese for a steadier first-use experience.
v1.5.0
- Inline Subtitle Controls: Toggle original / translated text, adjust font size, and tune background opacity directly in the subtitle window with live preview.
- Remembered Window Layout: Subtitle window position, size, and background opacity are restored automatically the next time you open it.
- Real-time Recognition Improvements: Updated the default real-time recognition model and improved display settings persistence for steadier long-running sessions.
v1.4.0
- Microphone Monitoring: Added clearer microphone input monitoring so users can quickly confirm whether audio capture is working.
- Translation Performance Improvements: Improved real-time translation responsiveness and stability, especially during longer sessions.
- Windows Installer Refresh: The website now provides the
1.4.0Windows installer for the latest download experience.
Highlights
- Windows Stable Release: The Windows installer is now officially available for Windows 10/11 (64-bit) users.
- 1.3 Architecture Upgrade: The core pipeline now separates recognition and translation, improving subtitle consistency in long-running sessions.
- Translation Engine Enhancements: Added DashScope translation integration with API key support and customizable prompts for better control.
- Cantonese Quality Improvements: Improved handling for Cantonese mixed Chinese-English input and auto-detected Cantonese scenarios, with final translations persisted in session history.
- Permissions and Stability: Refined microphone and macOS screen recording permission flows with post-authorization fallback detection, plus clearer audio capture and connection error reporting.