Release Notes

Bug Fixes

• Fixed a crash on startup in v0.10.12 (ImportError: could not import module 'PySide6.QtPrintSupport')
• All v0.10.12 users are strongly recommended to update to v0.10.13

New Features

• Added in-app notifications: the "Notifications" button in the toolbar shows an unread count, so announcements and support replies are now delivered inside the app (with dark mode support)
• Added "About antivirus warnings" link in the Help dialog (opens the official guide in your browser)

Improvements

• Reduced installer size

Bug Fixes

• Fixed real-time transcription failing to start (cases showing "Model not found" or "Backend not available")

Improvements

• Colored the transcription start button for better visibility
• Reduced false positives from antivirus software (embedded file properties into the executables)

Bug Fixes

• Fixed summarize window failing to open
• Fixed help screen content not displaying
• Fixed crash during app auto-update

New Features

• Added in-app Feedback button (toolbar) to send bug reports, feature requests, and questions directly from the app
• You can now preview exactly what diagnostic information will be attached before submitting feedback

Improvements

• Improved network connection reliability when sending feedback

Bug Fixes

• Fixed OpenVINO cache detection to use per-device manifest, resolving repeated "first-time optimization" dialog
• Fixed cache warning dialog appearing multiple times during batch processing
• Improved ETA estimation accuracy

Improvements

New Features

• Inline display of thinking, tool execution, and response sections in LLM chat
• First-run model download (Gemma 3n E4B Q4_K_M)

Improvements

• Significantly improved token counter accuracy

New Features

• OpenVINO encoder integration — Auto-download OpenVINO encoder for Intel GPU acceleration. Download option added to ModelHub

Bug Fixes

• Backend fallback fix — Fixed false-positive error detection that incorrectly stopped working backends (OpenVINO, Vulkan, CPU)
• Explicit backend protection — When a backend is manually selected, auto-fallback is now properly disabled

Bug Fixes

• Fixed 3 bugs in the license dialog
• Updated Welcome screen feature list to match latest content
• Fixed file lock issue during installer update

Improvements

New Features

• Free trial — 7-day free trial with all features unlocked
• LLM model capability badges — Auto-detect tool calling & reasoning support, displayed as badges in model selector
• Model Hub launch button — One-click Model Hub access from Settings
• Real-time transcription floating — Detach the real-time panel as an independent window

New Features

• Speaker diarization engine overhaul — Migrated to sherpa-onnx engine. CUDA / CPU support, per-model optimal thresholds for improved accuracy
• Real-time transcription — High-accuracy real-time transcription powered by Silero VAD. Dual capture (mic + internal audio), per-segment language switching, context carry-over
• App auto-updater — Automatic update notification and installation on new releases
• File association — Double-click audio files to open them directly in WhisperApp