superwhisper: vanhemmat versiot ja versiohistoria

Features and Fixes
- Fix microphone selection not saving between app restarts
- Resolve spoken self-corrections in Message preset
- Security fixes

Improvements
- Improved S1 connection error handling and fallback
- Improved paste reliability + notification when attempting to paste into apps running as administrator
- Use compression when sending large vocabulary lists with cloud requests
Fixes
- Fix company license activation link redirect in onboarding
- Fix `zh-TW` language normalization

Improvements
- Cohere is now the default local model on non ARM machines
- Download progress UI for model dependencies in the Models library
- Press Enter to save when editing a speaker names
Fixes
- Reprocessing a recording with a different mode no longer changes your active mode

## Features
- Windows context menu integration "Open with..." and Show more options "Transcribe with Superwhisper"
- Extended support for `.opus` files
- Whisper models are now free
- Free Pro trial now counted in words instead of minutes
- S1-mini is now generally available
- Tone control for Message, Mail, and Notes modes
- In-app updates for local models
## Improvements
- Faster launch and time to transcription-ready
- Whisper models renamed to Tiny, Base, Small, Medium, and Large

## Features
- Whisper models are now free
- Free Pro trial now counted in words instead of minutes
- S1-mini is now generally available
- Tone control for Message, Mail, and Notes modes
- In-app updates for local models
## Improvements
- Faster launch and time to transcription-ready
- Whisper models renamed to Tiny, Base, Small, Medium, and Large

## Features
- All-time stats view with X share card
- Updated S1-Mini model with tone control
- Cohere Transcribe is now marked as stable
- Word-by-word playback highlighting in History
- S1 announcement modal
## Improvements
- Significant performance boost for local Whisper models
- Vocabulary support for local Whisper models
- Updated application icon
- Smoother multipart model download progress updates
## Fixes
- Elevenlabs Scribe real-time "no voice found" fix
- Super mode parsing fix to remove errant `` tags in responses

## Features
- Added Cohere Transcribe model with Vocabulary support via our new Supervocab engine
- Added sound effect theme selection
- Added proxy support (WPAD and PAC)
- Added `record/start` and `record/stop` deep links for controlling recording from external tools
## Improvements
- Model selector now has arrow-key navigation
- Models that your system can't run are no longer available to download
- Updated the history info drawer UI
- Updated the reprocess button and its popout
- Experimental feature icons now sit beside their toggles and light up when active
- Added hover states to the vocabulary input buttons
- Improved model dependency system
## Fixes
- Fixed Settings window drag in history item full-view
- Fixed "popular" model options
- Fixed history transcript copying (full, segmented, and speaker-formatted output)
- Fixed sound effects toggle not being respected at launch
- Fixed the help message icon position
- Fixed escape key navigation when in history item full-view with info drawer open
- Fixed history info drawer timestamp format
- Fix cohere auto language detection

New Features
- Run language models fully on-device with the new experimental Superwhisper S1-mini model — no cloud required
- New (experimental) Parakeet multi language real-time model for live transcription across more languages
- Automatic fallback to raw transcription if a language model fails, so you never lose a recording
- New "Supervocab" vocabulary engine for local whisper models
Improvements
- Faster microphone initialization between recordings
- Mode keyboard shortcut moved out of Advanced settings for easier access
- Improved model library table layout
- Smooth transitions between "AI Processed", "Original" and "Segmented" history item views
- Language picker search input

New Features
- Use Opus audio format with compatible voice models for substantially faster cloud transcriptions
- Per-mode auto-capitalize insert / smart capitalization toggle
- Support for streaming language model responses
- Region routing for cloud transcriptions — requests are now routed to the lowest-latency region for faster turnaround
Improvements
- Implement smooth scrolling for long transcription results in recording window
- History list performance enhancements
- Make vocabulary input inner buttons clickable
- Superwhisper S1 transcriptions now display word-level segments correctly in the history view
Bug fixes and cleanup
- Migrate user config files from Roaming to Local in prep for Filesync feature implementation
- User can now use enter key to close/accept error windows
- Fix per-mode shortcut triggering
- Fix per-mode shortcut active mode revert issue

Features and Fixes
- Support for Parakeet local models on ARM based processors
- Models Library view for browsing, downloading and creating your own custom models

Features and Fixes
- Mode specific keyboard shortcuts
- All model files downloaded from Superwhisper host
- Implement license email change requests
- Move dynamic normalization to sound settings
- Mode change keyboard shortcut helper at bottom of modes list
- Advanced settings UI rework
- Add Opus 4.6 and 4.7 to custom model setup
- History search and multi select UI rework
- Fix bug with disabled but still clickable speaker separation toggle

Features and Fixes
- Vocabulary UI redesign
- Fix modes UI load time issues
- Add support for Azure Foundry custom model endpoints
- Crash fixes

Features and Fixes
- Modes UI overhaul
- Activation Apps UI and indexing overhaul
- Clickable timestamps in History segment view
- Model favorites UI
- Multiple crash fixes
- Fix custom Anthropic model creation flow
- Fix crash on drag recording window between monitors

Features and Fixes
- Toggle for "Automatically increase microphone volume" for improved voice detection
- Toggle for "Silence removal" to improve accuracy and reduce hallucinations
- Add mouse button shortcut for recording toggle and push-to-talk
- Add clipboard behavior preference to allow clipboard bypass and restoration
- "Identify Speakers" feature for Whisper, Parakeet Multi-language and most cloud models, including speaker naming UI
- SV-1 and SL-1 are now model defaults
- Context-menu on Settings -> Modes items right-click with duplicate, delete and set active
- History full-window view for long transcriptions
- Progress indicator for file decoding and transcription progress (applies to Whisper and Parakeet models)
- Elevenlabs Scribe V2 model with real-time support (currently available as an experimental model)
- Complete overhaul of onboarding flow with microphone check and default model selection
- License UI overhaul with improved deeplink handling

Features and Fixes
- Improved space insertion logic
- Add VAD state reset, should resolve repeated "no voice found" errors
- Add error notification for proper noun extractor initialization
- Add error notification for tray icon initialization
- Add quit option to recording window context menu

Features
- Parakeet English only model with real-time support added
- Global playback behavior config — set a default that can still be overridden per mode
- Brighter UI throughout the app
- Re-styled keyboard shortcut inputs for a cleaner look
- Removed push-to-talk toggle — clear the shortcut to disable push-to-talk instead
- Sound effects volume control added to settings
- Enhanced audio engine is now the default
- Recording window right-click context menu with Settings and Resize options
- Discard recording confirmation when cancelling recordings longer than 30 seconds
- Alert with shortcut to control panel shown when no audio inputs are available
- Improved activation app indexing performance
- Improved recording error notifications
- Faster initialization
- Faster session cancellation
Fixes
- Fixed recording window freeze on drag bug
- Fixed unclickable expand icon when "Start Recording" is shown
- Fixed stuck key scenarios
- Fixed audio input stream setup with a fallback mechanism
- Fixed recording window system close bug
- Fixed unable to delete pro models after trial expiry
- Added crash handlers for inaccessible config files
- Improved file resampling performance
- HOTFIX: v1.3.3 -> 1.3.4 Onboarding not displaying shortcuts