Superwhisper Release Notes

Features

You can now transcribe .mov video files directly from the file picker.

Vocabulary management is now simpler with a combined search and create input bar.

Experimental S1-mini model and Tone settings for more control over your transcriptions.

Fixes & Improvements

Updated vocabulary focus styling with a brightened stroke for better visibility.

Tightened the menu bar icon for a cleaner, narrower appearance.

Refined the typing speed test and stat sharing controls.

Polished the onboarding experience with smoother toast animations and improved card usability.

Fixed the Return key not releasing properly after closing the mode panel.

Resolved an issue affecting app framework packaging.

Improvements

• Improved the audio input device picker to favorite and exlude audio devices.
• New installations now use ~/superwhisper instead of ~/Documents/superwhisper as their configuration folder. Existing installations are unchanged, but we recommend switching your directory

Fixes

• Fix for handling control characters in some languages
• Additional safeties and recovery if app folder access is temporarily denied.
• Fixed for the mini window stealing focus from other apps if mode switching during dictation.

Features & Improvements

Added new deep links to start and stop recording, making it easier to control superwhisper from external shortcuts and automations.

Vocabulary hints for adding or replacing words are now clickable for quicker editing.

Improved sound effects with cleaner audio, now sounds good even at high volume on studio monitor headphones

Fixes

Fixed cloud transcription occasionally hitting API rate limits, improving reliability during longer sessions.

Fixed an issue where unexpected Unicode characters could appear in Japanese transcripts.

Features & Improvements

Faster pasting — time to paste latency reduced by 300ms

Improved transcription accuracy with enhanced vocabulary support and forced alignment (Whisper offline models)

Tightened history card UI with new expand/collapse reveal animations

Polished mini-mode pill with refined animations and tooltips

Cleaner sound effects — removed the trailing noise from sound effects

Updated DMG installer background and window layout

Now requires macOS 14 or later

Fixes

Fixed first realtime transcription occasionally failing after the app had been idle

Fixed duplicate messages appearing when dismissing multi-request agent sessions

Fixed recording clip height and collapse icon positioning in the recording view

Agent Escape/Enter shortcuts now respond only when the relevant window is focused

Fixed agent chat sizing issues, including dead space and table spacing

Resolved a freeze that could occur when displaying notifications

Features & Improvements

Added a copy button to code snippets in agent responses for quicker reuse.

Polish and refinement of Coding agent panel, beautiful animations and presentation

Refined a range of visual and interaction details across the share card, Super picker, top bar, and recording details views.

Fixes

Fixed words occasionally being dropped at the start of a real-time transcription session.

The recording window now stays clamped to the screen edge instead of jumping off-screen.

Stopped the agent window from drifting leftward when resizing.

Pressing Esc twice now only dismisses the agent window when it's visible.

Fixed issues with the Claude Code integration

Features & Improvements

Added theme control to switch between light, dark, and system appearance modes

New sound effect style & options

Experimental access to S1-Mini local model

Enhanced vocabulary processing - support for up very large vocabulary lists

Added Stat share card feature

Fixes

Meeting mode now respects language selection in AI summary

Model picker now shows checkmarks for downloaded models

Fixed "No voice found" error that could occur during live transcription

Improved S1 region failover and host handling reliability

Fixed side sheet and mode panel slide animations

Settings sidebar now resets to list view when reselecting tabs

Massive improvements to the local language model output quality

Features & Improvements

You will now get a warning when no audio is detected from your microphone
Performance improvement for all Cloud hosted voice models
Improved vocabulary and word timestamps handling on S1
Improved silence removal in noisy environments

Fixes

Fixed icon compatibility on macOS 13.3
Fixed CPU usage spike that occurred after extended use
Improved performance when transcribing from audio files
Fixed potential crashes when interacting with vocabulary items
Improved waveform display performance

Features & Improvements

Added GPT 5.5 on Bring Your Own Key

Enhanced tooltips in model and preset pickers with warm-state information

Added direct coding agent plugin installation capability

Implemented standard search functionality in History view

Added CSV import functionality for vocabulary with newest-first sorting

Enhanced audio processing with improved dynamic normalization and live waveform gating

Added Codex hook support for extended functionality

Fixes

Fixed recording window position drift across show mode selector cycles

Resolved stale chat panel state when advancing queue

Fixed queued message display issues

Improved recovery from realtime connection drops

Fixed mode change view closing behavior during recording cancellation

Restored missing questions functionality in Claude agent

Enhanced gain recovery for whisper-level recordings

Added success messaging for plugin installations

Features & Improvements

• Added Enterprise login support for business users
• Introduced per-mode keyboard shortcuts for Push-to-Talk and toggle recording
• Redesigned model library with improved table layout and chip colors
• Added Help menu with quick links to documentation and support
• S1 Voice & Language features are now stable (no longer experimental)
• Improved clipboard handling to prevent pollution of clipboard managers
• Enhanced license page design and user experience

Fixes

• Fixed crashes during long recordings and system audio meetings
• Resolved Parakeet model download hangs and failures
• Fixed missing first words in Scribe real-time transcription
• Improved transcription error handling and stability
• Fixed language model selection issues
• Resolved smart capitalization problems in Superhuman
• Fixed audio device selection when device list changes
• Prevented crashes when quitting the app
• Cleaned up email prompt greetings and hallucination detection

ElevenLabs Scribe v2

Scribe v2 + Scribe v2 realtime is now available to all users!

Features & Improvements

• New recording retention setting — automatically clean up old recordings on a configurable schedule
• Added lower volume option that reduces output volume to 15% while recording
• Per-mode autocapitalize toggle — disable auto-capitalization for specific modes
• Search-scoped select-all and copy in history view
• Filler words (uh, um etc) are now removed automatically from transcripts
• Added mode activation to mode card context menu
• Strip stray think/thought prefix lines from LLM output
• Fix british english language when used with Nova realtime

Performance

• Optimized history list faster search, navigation and selection

Fixes

• Fixed recording results not displaying vocabulary replacements and post-processing cleanups
• Fixed flash attention crash on newer Macs by updating GGML to 0.9.7 with automatic fallback
• Fixed mirror-mode microphone auto-switching bug caused by transient display states
• Fixed mute option ordering so playback suppression starts before recording begins
• Fixed LLM result handling when using S1 models

Realtime Streaming

• New realtime streaming UI with live transcription display
• Improved Parakeet streaming performance

Features & Improvements

• Added British English spelling transform for transcription
• New input device auto-selection
• Added search field to language model dropdown
• Added Report Issue option to history items
• Play pre-stop sound before stopping recording
• Speakers and segments are now respected when copying transcriptions
• Added anti-hallucination parameters to faster-whisper transcription
• Clearer cancel recording confirmation dialog labels

Fixes

• Fixed recording bottom bar layout with long hotkey labels
• Fixed crash report boot loop caused by corrupt database rows
• Fixed SIGABRT crash in GGML backend
• Fixed long recordings stuck on pending due to continuation leak
• Fixed negative time saved showing on homepage stats
• Fixed vocabulary hints breaking ElevenLabs Scribe transcription
• Fixed crash when duplicating or deleting modes
• Fixed realtime streams not opening mid-recording
• Fixed websocket close crash on nil socket
• Normalized language codes for ElevenLabs WebSocket connections
• Fixed chevron rotation direction for model provider groups

ElevenLabs Scribe (experimental)

• Added ElevenLabs Scribe as a new cloud transcription provider (punctuation issues known, use at your own risk), enable experimental models in advanced settings