Changelog

New

• Audio Enhancement: local AI reduces background noise and room reverb in dictation and meeting recordings, plus speaker echo in meetings. Enabled by default; configurable in Settings → Mic & Audio.
• More local models: added 13 speech recognition models, including additional Parakeet variants, Cohere, SenseVoice, Paraformer, and Nemotron.
• Help & Feedback: a new sidebar menu brings together support, feedback, the feature tour, release notes, updates, and community links.
• System audio controls: toggle system audio for the current call directly from the recording bar, with a clear crossed-out speaker when it is off.

Improved

• Improved memory use, cancellation, and interrupted transcription recovery for long recordings with Parakeet and Whisper Turbo / Turbo Q5.
• More stable transcription progress, with a single progress indicator and Cancel control at the bottom of the meeting window.
• Clear warnings when system audio capture is interrupted, unavailable, or missing permission.
• A more compact microphone priority list keeps audio settings easier to reach, with scrolling for additional devices.

Fixed

• Fixed system audio remaining disabled when starting another call or explicitly continuing a recording.
• Fixed starting Retranscribe on another recording cancelling an ongoing transcription.
• Fixed imported recordings remaining stuck in transcription or showing an empty transcript after processing finished.
• Fixed transcription requests being incorrectly routed when using supported third-party cloud providers.
• Fixed copy and paste shortcuts with Dvorak keyboard layouts.
• Fixed dropdown options being clipped near the bottom of settings windows.
• Fixed Ottex Camera appearing in the Dock as a foreground application.

New

• Agent-assisted configuration: external AI agents can configure dictation profiles, dictionaries, snippets, AI shortcuts, meeting projects, and templates.
• Editable configuration files: changes apply without restarting. Invalid edits leave the last working settings active.
• Archive and restore: disable profiles, shortcuts, projects, templates, and prompts while retaining their settings.
• Google AI Studio live transcription support: Smart and Verbatim modes are available with a Google AI Studio API key.

Improved

• Copy AI chat messages in meeting chats, standalone chats, and the Omnibar.
• Reduced interface overhead when browsing meetings, searching transcripts, and opening meeting details.
• Improved subscription plan comparisons.

Fixed

• Fixed transcript loss when continuing Soniox recordings.
• Fixed premature completion before the final transcript was saved.
• Fixed AI shortcuts selecting unavailable providers despite an available alternative.
• Removed an unrelated calendar-link example from voice-snippet instructions.

New

• AI-powered meeting notes: ask the AI to create or update notes directly in your meeting. Turn the conversation into decisions, action items, or a summary without copying text from chat.
• Automatic chat follow-ups: keep the AI working as your meeting progresses. Set a repeating instruction, such as updating action items every two minutes, using the latest conversation. Available in meeting and free-form chats.
• Camera Watermark (Experimental): show a Recording indicator on your video in Google Meet or Zoom, with optional alerts when Ottex Camera is not active. Enable it in Settings → Experiments, then select Ottex Camera in your meeting app.

Improved

• More ChatGPT models: added GPT-6 Astra, Sol, and Luna.
• A more readable recording timer for meetings lasting over an hour.
• Clearer transcription errors and retry options, with saved audio available for recovery.

Fixed

• Fixed Calendar returning to the Connect screen after access had already been granted.
• Fixed join reminders appearing for meetings you were already recording.
• Fixed long messages being cut off in AI Chat.
• Fixed truncated text in the Delete Project confirmation.
• Fixed temporary access problems with saved sign-in details being incorrectly reported as a sign-out.
• Fixed failed meeting transcriptions being shown as completed with no speech detected.

Improved

• Bug fixes and general reliability improvements.

Improved

• Bug fixes and general reliability improvements.

New

• Integrations (Experimental): connect Notion, Linear, ClickUp, Todoist, and Attio, then ask the Ottex agent to create tasks, update documents, or turn meeting notes into follow-ups. Enable Integrations in Settings → Experiments to try them.
• Project webhooks: automatically send meeting data after post-meeting templates finish. Configure multiple endpoints, custom headers, and included fields, with test requests and automatic retries.
• Multiple automatic templates: generate several notes and summaries after each meeting, with separate speaker identification and per-project settings.
• Open audio from Finder: use Open With → Ottex to open an audio file in the import dialog.

Improved

• Improved live transcript performance and stability during long meetings.
• More reliable storage for large dictation histories.

Fixed

• Fixed transcript editing not ending correctly after focus changes and duplicate saves when pressing Escape.
• Fixed message editors losing their width when entering edit mode.
• Fixed scrolling and layout issues in the meeting AI chat when using the @ mention menu.
• Improved error handling for sign-in, unavailable AI providers, and transcription retries.

Improved

• Long meeting transcripts now scroll much more smoothly, with significantly lower CPU usage and fewer interface freezes.
• Meeting prompts and AI model options now load without blocking the interface.

Fixed

• Fixed an issue that could omit the final words of a realtime transcription when stopping a recording.
• Fixed transcript text being clipped or incorrectly sized while editing.
• Fixed delayed speaker-name updates during live meetings.
• Fixed several smaller issues affecting long transcripts, AI provider availability, prompt loading, and meeting stability.

Improved

• Improved live transcription performance and stability, especially for longer meetings with multiple speakers.

New

• Google AI Studio: added support for Gemini 3.5 Transcribe Live.
• Ottex Cloud: Gemini 3.5 Transcribe Live is now the default dictation model.

Fixed

• Fixed an issue where signing in during onboarding could leave the Default profile pinned to the legacy Voice Balanced model. Affected profiles are repaired automatically.
• Fixed several smaller issues affecting onboarding, model selection, and transcription reliability.

New

• People and speakers: merge duplicate profiles more easily and manage speaker identities directly from meeting transcripts.

Improved

• More reliable meeting and dictation audio while other apps are using the microphone, with better recovery after audio-device changes.
• Smoother live transcripts and a more responsive meeting screen during long recordings, including safer editing while a meeting is in progress.
• Automatic recovery from temporary network interruptions and interrupted meeting processing, while local audio continues to be recorded safely.
• More consistent calendar participants, speaker assignments, Context Awareness, menus, text editing, and Settings.

Improved

• Bug fixes and general reliability improvements.

Improved

• Bug fixes and general reliability improvements.

New

• Automatic Meeting Summaries: Ottex now creates a structured summary after a meeting ends and opens it automatically when it is ready.
• Meeting summaries can use participant information to identify speakers while preserving any names you assigned manually.

Improved

• More stable microphone audio when other apps are using the microphone, including Google Meet and similar meeting apps.
• Improved recording reliability with AirPods, including recovery when the microphone unexpectedly becomes quiet.
• Smoother live transcripts and a more responsive meeting screen during longer recordings.
• Reduced CPU and memory usage during meetings, especially during long recording sessions.
• Meetings with little or no recognized speech now finish and save normally instead of appearing to fail.
• Faster and more reliable recording startup, with the beginning of your speech captured more consistently.
• Reduced unwanted changes to music volume and sound quality when dictation starts.
• Short recordings now stop cleanly without leaving Ottex active in the background.
• Reduced CPU usage while Ottex is idle and when moving the cursor between displays.
• Cleaner AI Assistant responses with fewer unwanted formatting artifacts.
• More reliable automatic model selection and fallback when the preferred model is unavailable.
• Improved overall recording recovery and stability after audio-device or system configuration changes.

Fixed

• Fixed occasional stuttering in meeting loading indicators and live transcript updates.
• Fixed repeated speaker processing that could slow down longer realtime meetings.
• Fixed several smaller issues affecting meeting recording, transcription, audio devices, and AI processing.

New

• Mouse Button Hotkeys: assign a mouse side button to Toggle Recording or Push-to-Talk, keeping dictation within reach without using the keyboard.
• Deepgram Privacy Mode: choose models that opt your transcription requests out of Deepgram's Model Improvement Program.
• Meeting and free-form AI Chat responses now appear as they are generated, so you can start reading immediately.
• Stop a long AI response at any time while keeping the useful part already generated.

Improved

• Automatic recording stop when your meeting ends, preventing forgotten recordings from continuing in the background.
• A clear warning when only your microphone is being recorded, helping you avoid meetings where the other participants are missing from the recording.
• More reliable audio recovery after an interrupted finalization or app restart, reducing lost recordings and missing-audio errors during retry.
• Clear explanations and direct next steps when a meeting cannot start because you are signed out or its transcription model is unavailable.
• Multiple email addresses for each person, making the same participant easier to recognize across calendars and meetings.
• Safe duplicate merging without losing meeting history, speaker assignments, aliases, or voice identity.
• More accurate company grouping without unwanted company entries for personal email providers such as Gmail or iCloud.
• Deepgram Provider: added Flux English and Flux Multilingual for responsive realtime transcription with conversational turn detection.
• Clear separation between realtime and pre-recorded Deepgram models, making it easier to choose the right option for live dictation or imported audio.
• Clearer model pickers with Real-time and Pre-recorded categories, capability badges, and automatic model selection.
• More reliable menus, dropdowns, and tabs that fit within the available window and keep focus in the expected place.
• Smoother text editing across meeting transcripts, AI Chat, and Settings.

Fixed

• Unexpected sign-outs during temporary network or authentication-provider outages.
• Occasional transcript sharing failures accompanied by an unclear empty-segment error.
• Inconsistent behavior across meeting controls, AI Chat, hotkeys, and Settings.

New

• Mouse Button Hotkeys: assign a mouse side button to Toggle Recording or Push-to-Talk, keeping dictation within reach without using the keyboard.
• Deepgram Privacy Mode: choose models that opt your transcription requests out of Deepgram's Model Improvement Program.
• Meeting and free-form AI Chat responses now appear as they are generated, so you can start reading immediately.
• Stop a long AI response at any time while keeping the useful part already generated.

Improved

• Automatic recording stop when your meeting ends, preventing forgotten recordings from continuing in the background.
• A clear warning when only your microphone is being recorded, helping you avoid meetings where the other participants are missing from the recording.
• More reliable audio recovery after an interrupted finalization or app restart, reducing lost recordings and missing-audio errors during retry.
• Clear explanations and direct next steps when a meeting cannot start because you are signed out or its transcription model is unavailable.
• Multiple email addresses for each person, making the same participant easier to recognize across calendars and meetings.
• Safe duplicate merging without losing meeting history, speaker assignments, aliases, or voice identity.
• More accurate company grouping without unwanted company entries for personal email providers such as Gmail or iCloud.
• Deepgram Flux English and Flux Multilingual for responsive realtime transcription with conversational turn detection.
• Clear separation between realtime and pre-recorded Deepgram models, making it easier to choose the right option for live dictation or imported audio.
• Clearer model pickers with Real-time and Pre-recorded categories, capability badges, and automatic model selection.
• More reliable menus, dropdowns, and tabs that fit within the available window and keep focus in the expected place.
• Smoother text editing across meeting transcripts, AI Chat, and Settings.

Fixed

• Unexpected sign-outs during temporary network or authentication-provider outages.
• Occasional transcript sharing failures accompanied by an unclear empty-segment error.
• Inconsistent behavior across meeting controls, AI Chat, hotkeys, and Settings.

New

• A dedicated AI Chat panel inside every meeting, keeping questions and answers alongside the meeting context.
• GPT-5.6 Luna, Terra, and Sol in the ChatGPT provider, giving you more model choices for AI features.
• New Deepgram transcription models and speaker diarization for clearer speaker-aware meeting transcripts.

Improved

• Settings are clearer and better organized.
• Meeting preferences now live in the main Settings area, making them easier to find.

Fixed

• Meeting recordings that could sound unexpectedly quiet.

New

• Ask Meeting AI Chat questions using the transcript, notes, summary, and meeting context.
• Meeting prompts, People, and Companies make follow-ups faster and meetings easier to navigate.
• Notifications for calendar events and automatically detected meetings help you start recording on time.

Improved

• More consistent meetings, recording, chat, prompts, participant controls, and interface behavior.

Improved

• Bug fixes and general reliability improvements.

Improved

• Improved performance and transcription speed.

New

• Major upgrade to Meetings with realtime recording, diarization, speaker recognition, and smarter meeting transcripts.
• Improved meeting flows with calendar-aware recording, better speaker handling, search, templates, and exports.
• Added public meeting sharing so selected transcripts can be shared with others.
• Added language selection and language prompt support for transcription models.
• Added support for new speech models, including Soniox async v5 and AssemblyAI Universal-3 Pro.
• Added Pro and Team subscription support with organization management.

Fixed

• Fixed meeting recording audio pipeline issues, including realtime audio, AirPods capture, and system audio changes.
• Reduced CPU usage during meeting recording.
• Fixed a bunch of minor bugs.

Improved

• Improved Settings with a resizable window, built-in update controls, and cleaner model picker layouts.
• Improved hotkey controls for mode-specific workflows and Scribe quick notes.
• Polished microphone, model, and recording widget states, including profile and live-preview visibility.

Fixed

• Fixed account and model refresh issues that could make unavailable-model errors harder to recover from.
• Fixed Scribe, calendar, transcript editing, and search UI edge cases.

Improved

• Improved app update and download plumbing, including release metadata used by Homebrew installs.
• Minor macOS UI polish and reliability improvements.

Fixed

• Fixed small update-flow and distribution edge cases.

New

• ElevenLabs support for Ottex Provider — use ElevenLabs speech models through your Ottex account.
• Transcript cleanup policy — choose how long to keep transcripts, set Keep Last N, and keep starred transcripts from being deleted automatically.
• Trim Silence toggle in Audio settings — remove silent portions from recordings before sending them to the cloud to reduce file size and API costs.

Improved

• Updated app branding — refreshed the Ottex app icon, logo, and theme polish.

Fixed

• Fixed high CPU usage during recording by throttling live microphone meter updates.
• Fixed streaming models dropping the first few words when speaking immediately after starting a recording.
• Fixed Ottex app navigation jumping back to Home after switching away and returning to the app.

Fixed

• Fixed the floating recording widget so Stop and X stay clickable without stealing focus
• Improved login and billing success flow with a more reliable fallback back into the app

New

• Added quick copy in the menu bar

Fixed

• Fixed media pause-start bug during recording start
• Fixed omnibar AI shortcut not returning focus to the previous app
• Fixed the "Open App" button resetting the current app view

New

• Auto-mute music — automatically pauses music and video playback when you start recording, resumes when you stop. Works with Spotify, Apple Music, YouTube, and any media app.

Fixed

• Fixed AI shortcuts not pasting results back into the original app
• Fixed a crash when scrolling through transcription history

Improved

• Better error messages when a profile's AI model is unavailable — with a direct link to fix it
• Profile selector stays visible during errors so you can switch models without leaving the widget

Fixed

• Widget buttons now respond to the first click instantly
• Fixed profile tooltips not displaying correctly in some edge cases

New

• Ottex AI Provider — access all AI models with a single account, no API keys to manage
• Realtime streaming transcription via Deepgram, Soniox, and Mistral
• Profiles Marketplace with 20 pre-built profiles (email, coding, medical, legal, etc.)
• New Models settings page
• macOS Tahoe compatibility fixes

Improved

• Onboarding simplified — fewer steps, cleaner dialogs
• Audio diagnostics and device resolution logging

Fixed

• Fixed sidebar hit-testing on macOS Tahoe

• Fixed Mistral AI provider compatibility with Voxtral Transcribe 2

• Local AI Models — transcribe entirely on-device with Parakeet V3, Whisper Large V3, Qwen3-ASR, and more. No internet required.
• Meeting Recording & File Transcription — record calls with system audio capture and speaker diarization. Import and transcribe audio files.
• Mistral AI Provider — use Voxtral Transcribe 2 for fast, accurate audio transcription.
• Custom OpenAI-Compatible Providers — connect any OpenAI-compatible API with your own endpoint and key.
• Improved Audio Reliability — rebuilt audio engine with better Bluetooth support, automatic mic reconnection.
• New Onboarding — redesigned setup experience with guided microphone testing and voice practice.

• Added support for Whisper series models
• Per-Mode Voice and Post-Processing Model Override — choose different transcription models for specific apps and websites
• New Providers: Groq, Fireworks AI — ultra-fast Whisper transcription and OSS models
• Improved Omnibar application indexing
• Dictionary moved to separate tab

New

• Single key hotkeys — use Fn, Escape, or modifier keys alone for push-to-talk and cancel recording
• Context awareness per mode — override screenshot capture settings for individual instruction modes

Fixed

• Minor bug fixes and reliability improvements

• Introducing "App Modes" — create custom rules for how to format results in Gmail, Cursor, or any other app

• Command mode — select text, then ask Ottex with your voice to make edits for you
• Added Google AI Studio API provider
• Added optional hints field to AI snippets to help the AI correctly insert them into surrounding text

New

• Added extra push-to-talk hotkey

Improved

• Improved Bluetooth headphones reliability

New

• Renamed the app to Ottex

Fixed

• Fixed minor bugs with the screenshotting tool

• Redesign & new logo
• Improved Voice-to-Text quality
• Voice-to-Text settings with new features: punctuation, new lines, paragraphs, numbered lists, and meta commands

New

• Omnibar shows immediately without beeping sounds
• AI shortcuts: with no text selection, a textarea appears for manual input

Improved

• Multiple screenshooter improvements and bugfixes

• BYOK with OpenRouter API key
• Custom models support for AI Shortcuts
• Custom models support for Transcription

• Home screen with statistics and transcription history
• "Retry transcription" functionality

• Voice transcription with AI shortcuts
• Screenshot capture and editing
• Omnibar launcher (Cmd+Shift+Space)
• Auto-updates via Sparkle