• Recording no longer asks for Screen Recording. The first time you record, macOS asks for System Audio Recording instead. For the most reliable automatic meeting detection, allow Accessibility in Settings > General > Permissions. • MCP support! Other apps on this Mac can now read your AI chat, the live transcript, and the meeting in progress. Turn it on in Settings > MCP. It stays on this computer only, and it asks for a token before anything can connect. • AI chat replies format as you read them, instead of waiting until the reply is finished. • A recording no longer stops itself a couple of seconds after a meeting starts. • The end of the meeting is included in the summary. • Reprocessing a meeting keeps the recording and your previous notes until the new notes are saved. • A brief glitch in meeting detection no longer ends a recording that is still going. • Names and terms you care about keep their spelling in the live transcript, the saved transcript, and the summary. Each custom word now has an Also heard as field for phrases the transcriber often gets wrong. • You can set the order of your export rules. In Settings > Processing > Export Integrations, move rules up or down. The first matching rule is the one that runs, and a rule that can never be reached is flagged. • If your Mac goes to sleep during a recording, the recording is saved first. If the meeting is still going when you wake, a new recording starts. • If you are updating from an older version, the menu shows a one-time prompt to grant System Audio Recording before your next meeting. • You can stop an AI reply while it is still coming in. Closing the chat cancels it. • You can choose how long meetings stay in History. The default is 90 days, and removing one from the list does not delete its saved transcript. • MeetingMind now requires macOS 26. If you are on macOS 15, the version you have keeps working, and this update will not install.

- While recording, click the microphone icon in the menu bar to open or close AI chat. Clicking the timer still opens the main menu. - You can also save the chat into your meeting notes with the document button in the chat input (off by default, remembered between meetings). - The AI Chat button now appears only while you are recording, and the empty Mic and System meters stay hidden until a recording starts so the menu stays compact.

- Turn on the new document button in the chat input row and the whole conversation is saved into your meeting file as an AI Chat section. Off by default, and remembered between meetings. - While recording, click the menu bar icon to open or close the AI chat directly, no need to open the full menu first. The AI Chat button now also appears only while you are recording, instead of sitting greyed out. - The empty Mic and System level meters are now hidden until a recording starts, so the menu stays compact.

New • Pause and resume recording. A pause button now sits both on the menu bar recording indicator and inside the dropdown, so you can pause without opening anything. Pausing drops the audio instead of recording silence, so a break or interruption never lands in your meeting file or transcript, and the timer freezes while paused. A recording left paused for over an hour ends automatically and processes as usual. • Streaming AI chat replies. When your AI provider supports it, chat responses now appear as they are written, word by word, within a second or two instead of waiting for the whole answer. Providers that do not support streaming fall back automatically. • Optional API key for Local AI servers. If your local or self hosted server needs authentication, add the key under Settings, AI Services, Local AI and it is sent as a standard bearer token on every request. Fixes • Microsoft Teams no longer starts a recording when you just open Teams or click into a chat. Recording now begins only when an actual call is in progress. • Clearer AI chat errors that explain what actually went wrong and name the service that could not be reached, plus an automatic retry when a connection drops right after recording starts or stops. • Testing an Anthropic API key no longer reports a false error for a valid key. • Live Insights no longer credit other participants’ questions or commitments to you. • The Question to Ask quick prompt no longer repeats the same suggestion. • Your Name now appears only once in Settings and is used everywhere consistently. • Fixed a Settings Corruption message that could block the app from launching. • Clicking the menu bar icon while a recording is processing now opens the menu, and dragging audio files onto the icon keeps working after your first recording. • Max Tokens now goes up to 65,536, giving reasoning models the headroom they need for longer chat replies. • More reliable automatic recording for Zoom, including cases where Zoom hides its meeting windows, while adding safeguards so ringtones, chat chimes, and simply having Zoom open never start a recording. • Smoother live audio meters during automatically started recordings. • Updated on device transcription engines for better speed and reliability.

• Live Export (beta): Stream your running transcript to a file in your Obsidian vault while you meet, so your other tools and assistants can follow along through your existing sync. Toggle it from the AI Chat window during a meeting. Off by default. • Clicking an upcoming meeting now reliably opens the right Zoom, Teams, Google Meet, or Webex link, even when the invite wraps the link in brackets or quotes. It no longer downloads a logo image by mistake. • The AI Chat now knows who you are from your Profile, so "what did I just say" or "summarize my points" no longer mixes up your words with the other person's. • Much better speaker separation in one on one calls: your words and the other person's are no longer swapped in the summary, including ad hoc calls with no calendar invite. • The other participant's name now appears for one on one meetings and ad hoc calls even when the calendar gives us very little to work with. • Real participant names are now kept in the saved transcript instead of being replaced with Speaker 1 and Speaker 2. • The AI Chat window now remembers its size, position, and whether the transcript panel was open, and your scratch notes stay put when you close and reopen it mid meeting. • History now shows estimated time remaining counting down, instead of elapsed time counting up. • The menu bar stays responsive while a recording is being saved and processed, even on long meetings. • AI responses now have up to five minutes to come back, so slower local models finish summaries instead of failing. • Recording no longer gets stuck after a blocked or failed start. The next recording just works. • Manual context you add during a meeting now stays anchored where you added it in the live transcript. • OpenRouter no longer quietly defaults to a paid model. You pick your model before it runs. • The Full Transcript in saved notes now reads as one clean block and renders properly in Obsidian and other markdown apps. • Fixed summaries failing with certain OpenAI reasoning models, and long recordings no longer drop small pieces of audio during processing. • Plus stability and crash fixes, and polish across export rules, transcription model management, and tag editing.

This is a massive update, and it would not have happened without you. Huge thanks to everyone who shared feedback, reported bugs, and helped test along the way. Every item below was shaped by what you sent in. The free tier daily recording limit is now 2 hours, up from 45 minutes. That is enough room for daily standups, client calls, and most working meetings before needing to upgrade. New Features • Speaker Diarization now runs after post-processing for both Parakeet and WhisperKit, labeling each line as Speaker 1, Speaker 2, etc. Pick Pyannote Fast or Pyannote Accurate under Settings > Recording > Speaker Diarization. • Refreshed Settings window with a grouped sidebar (Workspace, Capture, Pro), color-coded icons, and inset rounded cards. • Live transcript replaces [You] with your real name from your Profile. One-on-one meetings also swap [Remote] for the attendee’s name; group meetings keep [Remote] generic so distinct speakers are not collapsed. • Meeting Info panel at the top of the AI chat sidebar shows the meeting title and remote attendees from the calendar invite. • Microsoft Teams auto start/stop now matches Zoom in reliability, including on macOS Sequoia and Tahoe where window titles are stripped. • Filename Format adds a {userTitle} token that fills in the title you enter at the meeting prompt. • Redeem Code button on the Purchase page for App Store promo and offer codes. • New Passthrough Live Transcript option for faster summaries when live transcription is reliable in your setup. Improvements • Max tokens slider now goes up to 16,384 (default 4,096), giving reasoning models like Gemini 2.5 Pro room to think. • Menu now surfaces clear banners when an AI Service is not configured or when Microphone, Screen Recording, or Notifications permission is denied, with deep links straight to the right pane. • Delete All button on the unprocessed recordings recovery panel. • Notion export retries on rate limits and server errors, and splits pages over 100 blocks automatically. • Parakeet v3 transcription is more accurate for non-English and mixed-script content. Bug Fixes • Fixed prior meeting attendees and speaker labels leaking into the next recording when the new meeting has no calendar match. • Live transcript no longer flashes or yanks you back to the bottom while you scroll up to review. Text now streams in character by character. • Fixed back-to-back meetings using the wrong title when a new meeting starts during processing. • Fixed ^context^ markers from AI chat occasionally missing from the saved markdown. • Live transcript strips Whisper non-speech annotations like (keyboard clicking) and (background noise). • Fixed the final words of live transcription being dropped when stopping mid-chunk. • Fixed export rules not firing because they matched against the generated filename instead of the meeting title. • Fixed reprocess from History Manager failing silently after a file was renamed during processing. • Fixed exported markdown in Obsidian and Notion losing their date stamp; Obsidian writes are now atomic. • Fixed already-processed recordings resurfacing in the unprocessed recovery panel on launch. • Fixed speaker-name inference for corporate “Last, First” calendar formats (Outlook, Exchange). • Improved purchase reliability: unverified transactions are cleared, refunds are honored immediately, premium gates re-check at execution time, Unlimited users no longer flicker to free on slow StoreKit launches, and a 15-second timeout prevents the Buy button from hanging. • AI Services tab no longer triggers Keychain prompts on every keystroke. • Restored the full AI summary structure, including Key Participants and Speaker Contributions. • Fixed the previous meeting’s summary appearing in Live Chat during the next recording. • Fixed processing items getting stuck in the queue forever after a quit-during-processing. • Fixed the Finish your Profile banner persisting after a name was entered.

Fixed Google Meet auto-detection for Chrome, Firefox Your Profile in Settings for accurate speaker attribution in AI summaries Recording quality settings with Low, Medium, High, and Lossless options Voice Activity Detection skips silence, reducing CPU usage and preventing hallucinated text Dual-stream transcription with automatic [You] and [Remote] speaker labels Transcript cleanup removes repeated phrases and filler words AI insights now track conversation state across the meeting for better suggestions Smart insight pre-filtering reduces unnecessary API calls by up to 90% "Question to Ask" quick prompt in AI Chat Searchable settings sidebar Comprehensive thread safety audit across the entire codebase Text-based acoustic echo filter for duplicate transcription suppression Microphone and system audio toggles locked during recording Reduced CPU usage during recording 25+ bug fixes including history entries disappearing when notes are relocated, bookmarked moments not appearing in markdown, and back-to-back meeting title handling

• Copy and select text from AI chat responses with new message action buttons • Confirmation dialog when quitting during active processing to prevent lost work • Improved reliability when saving recordings to cloud storage (iCloud, Google Drive, Synology) • Fixed audio sample rate mismatches between microphone and system audio with clear warnings • Better support for back-to-back recordings with correct meeting titles in the menu bar • AI insight analysis now gracefully backs off after API errors instead of retrying aggressively • Added 1-day memory retention option for AI conversation context • Fixed calendar permission sync when system authorization is revoked • Improved compatibility with GPT-5, o1, and o3 models • Warning for emoji characters in output folder paths that can cause issues • Various stability and performance improvements

New Features • Transcribe-only mode — export clean transcripts without AI summarization • Automatic recovery of unprocessed recordings after unexpected app closures • New keyboard shortcut ⇧⌥R to start/stop recordings from anywhere • Support for LM Studio, LocalAI, Jan, and other OpenAI-compatible local AI servers • History now stores up to 100 recordings • Keyboard shortcuts reference available in Settings Bug Fixes • Improved Notion export reliability • Better resource cleanup on app quit • Fixed speaker diarization adding incorrect spaces in transcripts • Fixed issues with meeting titles not appearing correctly in filenames • Fixed AI summaries being incorrectly stripped from output • Fixed rapid start/stop recording issues • Fixed race conditions in Obsidian export and keyboard shortcuts

Improved Microsoft Teams meeting detection with fewer false positives Fixed a crash that could occur when stopping recordings Better handling of screen share transitions during meetings Enhanced AI transcription reliability Various UI and stability improvements

Automatically identify and label speakers in transcripts when using Parakeet Customize recording filenames with new tokens like {YYYY}, {title}, and {platform} Automatically route meeting notes to Obsidian or Notion based on meeting names using smart export rules Get instant AI-generated answers when questions are asked during meetings, displayed as teal insights Generate a sanitized debug bundle in Settings to help our support team resolve issues faster Easily resolve overlapping calendar conflicts by choosing which meeting to associate with your recording Twice as organized with a redesigned Settings window featuring colorful icons and macOS styling Live Insights now appear in a collapsible left sidebar in AI Chat with a lightbulb toggle Reduced AI API costs by 80-90% with smart trigger detection that only analyzes key moments Rename recordings inline by double-clicking any entry in History Manager—all files update automatically Per-meeting session logs now capture detailed activity while automatically masking sensitive data Fixed "Buy Unlimited" button occasionally becoming unresponsive or showing incorrect pricing Fixed false Microsoft Teams meeting detection triggered by broad window titles Fixed insights sensitivity settings being ignored Fixed insights sometimes not appearing during meetings at startup Fixed Obsidian export rules not matching meetings with timestamps in filenames Resolved a crash when exporting settings containing binary data

This update focuses on rock-solid reliability for your meeting workflows. We've fixed critical issues that were preventing automatic recording from working consistently and resolved bugs that made importing audio files frustrating. If you've experienced any of these issues, this update is for you.

• Enhanced recording feedback: The menu bar now displays a "Saving..." animation when you stop a recording, providing clear visual feedback while your audio is being prepared for processing • Cleaner status bar design: Simplified the processing indicator by removing the item count for a more streamlined appearance • Fixed automatic recording for consecutive meetings: MeetingMind now reliably auto-starts back-to-back meetings without manual intervention. Previously, only the first meeting would auto-start correctly • Fixed manual audio import workflow: The "Add Recording" button in History Manager now works for multiple consecutive imports. Each import properly generates markdown summaries without permission errors • Resolved consecutive recording freeze: You can now start multiple manual recordings in a row without experiencing any delays or application hangs

We've completely rebuilt MeetingMind's live transcription engine with cutting-edge technology that delivers dramatically better results: • 10x More Accurate Transcription – Upgraded to Whisper.cpp, the same technology powering professional transcription services. Your meeting transcripts will now capture conversations with exceptional accuracy, including technical terms, proper nouns, and complex discussions. • Lightning-Fast Response – Experience near-instant transcription with less than half a second of delay. Watch your words appear on screen in real-time as you speak. • Smoother Performance – Significantly reduced CPU usage and optimized update frequency means your Mac runs cooler and battery lasts longer during meetings. The app now intelligently adjusts its update speed based on whether you're actively viewing transcripts. Refined Menu Bar Experience • Cleaner Meeting Display – Simplified how active meetings appear in your menu bar by removing unnecessary time indicators. You'll now see just the meeting title without distracting countdown timers. • Rock-Solid Timer Display – Fixed an issue where the recording timer could wrap or resize unexpectedly. The menu bar status now maintains a consistent width across all your monitors for a professional, polished look. Under the Hood This update includes numerous technical improvements to ensure MeetingMind runs reliably and efficiently on your Mac. We've resolved several edge cases, improved thread safety, and optimized audio processing for better overall performance. Your Privacy Matters As always, all transcription happens entirely on your Mac. No audio or transcripts are ever sent to external servers during live recording. --- We're committed to making MeetingMind the best meeting assistant for Mac. Thank you for your continued support!

NEW FEATURES • Recording Tags - Organize your recordings with custom tags featuring 13 colors and 17 icons. Filter your history by tag to quickly find specific meetings. • AI Context Length Validation - Automatically detects if your AI server's context window is too small for meeting summaries, with visual warnings to help you configure it correctly. • Minimum Recording Duration - Recordings shorter than 1 minute are now rejected automatically to prevent wasted processing on accidental recordings. IMPROVEMENTS • Streamlined AI Model Selection - Updated to latest GPT-5 and Claude 4.5 models with easy dropdown menus. Defaults now set to the most recent versions (GPT-5 and Claude Sonnet 4.5). • Parakeet Transcription Engine - Now the default engine for 10-100x faster processing with multilingual support for 25 European languages. BUG FIXES • Fixed multi-monitor menu bar glitch where the status item would flicker when moving between displays with different resolutions. • Fixed settings export crash caused by JSON serialization issues with custom quick prompts. • Fixed transcription crashes during model loading with better validation and error messages. • Fixed overtime meeting titles - recordings are now correctly named using the calendar event instead of generic "Recording_XXXXX" names. • Fixed timer display briefly showing "00:00" when opening the menu during active recordings. • Fixed history item deletion for recordings that were renamed after processing.

Zoom Integration & Meeting Enhancements • Automatic Microphone Sync with Zoom - Your microphone now automatically mutes and unmutes in sync with Zoom's keyboard shortcut (⌘⇧A), keeping your recording perfectly aligned with meeting audio. See documentation in Settings. • Smart Calendar Filtering - Choose exactly which calendars to monitor for automatic meeting detection, and filter between all calendar events or just online meetings Recording & Processing Improvements • Recording Reprocessing - Regenerate summaries from existing recordings with updated AI prompts or settings - perfect for fixing incorrect summaries or trying new AI configurations • Parakeet Transcription Engine - New ultra-fast transcription option that's 10-100x faster than Whisper for post-processing, with support for 25 languages • Content-Agnostic AI Processing - Process any audio content beyond meetings: podcasts, YouTube videos, lectures, interviews, and more with AI-adapted summaries Enhanced History & Management • Improved History Manager - Remove items from history lists without deleting files, better organization of past recordings, and failed recording retry options (⌘⇧H) • Time Saved Dashboard - New usage statistics showing how much time you've saved on transcription and note-taking, with detailed breakdowns • Better File Organization - Standardized file naming and smarter handling of duplicate filenames User Experience Enhancements • Clearer Error Messages - See actual error details instead of generic messages, with helpful troubleshooting guidance • Settings Preservation - Usage statistics and onboarding state now preserved when resetting settings • Improved UI Alignment - Better visual consistency across all settings panels • Enhanced Settings Organization - Auto-recording settings moved to top of Recording Settings for better workflow Technical Improvements • Thread Safety Enhancements - Improved stability for USB microphones and live transcription on all macOS versions • Better Performance - Optimized audio processing and reduced memory usage • Enhanced SecurityAdded URL scheme validation for meeting links to protect against malicious calendar events. Only safe protocols are now allowed (http, https, zoommtg, msteams, meet, webex). • 24-bit USB Microphone SupportFixed compatibility with professional USB microphones like the Razer Seiren V3. These devices now work correctly with proper 24-bit audio format detection and conversion.

New Features • Parakeet Transcription Engine – Process your recordings up to 100x faster with our new alternative engine supporting 25 languages, powered by Apple's Neural Engine for incredible performance • History Manager – Browse, search, and chat with AI about any past meeting. Add previous meetings as context to your live recordings for richer AI insights • Usage Statistics Dashboard – Track your productivity with beautiful charts showing meetings recorded, total hours, files generated, and daily patterns at a glance • Smarter Error Messages – No more confusing generic errors. See exactly what went wrong with clear, actionable explanations when issues occur • Settings Protection – Import/export now validates compatibility between app versions, preventing configuration issues Improvements • Custom notification sounds for recording events - consistent experience across all systems • Smarter file handling when transcription fails - keep your original files for easy retry • Alphabetically sorted AI models for easier browsing • Cleaner error dialogs focused on what matters • Better error details with complete diagnostic chains • Smoother menu bar interactions with improved state management • Enhanced calendar integration with better error handling • Progress indicators for both transcription engines • Audio format validation catches issues earlier • Usage statistics included in settings backup Bug Fixes • Resolved crash in live transcription with USB microphones • Eliminated audio priority issues for smoother recording • Fixed model download status tracking • Improved memory management in transcription engine • Better thread safety throughout the app • Download tasks clean up properly when closing the app • Settings validation prevents crashes from corrupted preferences

NEW FEATURES • Drag-and-Drop Audio Processing: Drag M4A, WAV, MP3, AIFF, CAF, or AAC files onto the menu bar icon to transcribe and generate AI summaries - perfect for reprocessing failed recordings or analyzing external audio files • Launch on Login: Automatically start MeetingMind when you log into Mac with easy toggle in Advanced Settings • Apple Intelligence Integration: Native on-device AI processing using Apple's FoundationModels framework (macOS 26+) as automatic fallback - zero configuration, complete privacy, works offline • Enhanced Speaker Identification: AI now systematically tracks speakers with pre-processing pass, voice pattern recognition, and contextual attribution for more accurate meeting transcriptions • Configurable Markdown Transcript Output: Toggle transcript inclusion in markdown files to save disk space or include verbatim transcripts when needed • Custom Vocabulary System: Add domain-specific terms, acronyms, and specialized vocabulary with beautiful bubble interface - improves transcription accuracy and AI understanding for technical, medical, and company-specific language • Custom WhisperKit Model Support: Integrate any HuggingFace WhisperKit-compatible repository with real-time validation - enables testing cutting-edge, fine-tuned, or community-optimized transcription models IMPROVEMENTS • Three-tier AI service priority: Ollama (local) → OpenRouter (cloud) → Apple Intelligence (on-device) with automatic fallback • Apple Intelligence availability checking in Settings with real-time status indicators • Conditional compilation support for Apple Intelligence ensures backward compatibility with macOS 15 • Complete privacy for Apple Intelligence processing - no external data transmission, works offline • Drag-drop overlay preserves menu bar button clicks with intelligent passthrough behavior • Real-time progress tracking for dropped files with visual stage indicators • Sequential batch processing support for multiple dropped audio files • Format validation with helpful error messages for unsupported file types • Launch on login preference included in settings export/import for easy backup and transfer • AI speaker tracking with technical vocabulary and expertise indicators for role-based identification • Confidence level tracking for speaker attribution (confirmed, likely, insufficient data) • Cross-validation system retroactively updates speaker identity when named later in transcript • Contextual uncertainty handling with identifying characteristics instead of generic 'Unknown' labels • Simplified meeting detection with static 3-second polling interval for consistent, predictable behavior • Optimized AI model parameters for more consistent speaker labeling (temperature: 0.15, max tokens: 2500) • Real-time WhisperKit model download progress with visual stage indicators (downloading 30%, preparing 60%, loading 80%, ready 100%) • Custom WhisperKit model downloads persist across app sessions via UserDefaults tracking • WhisperKit repository validation with visual feedback (green checkmark for valid, red X for invalid repos) • OpenRouter model auto-discovery on settings view appearance eliminates manual Discover button • AI Services settings reorganization - markdown theme moved from Advanced Settings for logical grouping with chat preferences

NEW FEATURES • Unified AI architecture with dynamic model discovery - now works seamlessly with Ollama, LM Studio, LocalAI, Jan, and OpenRouter with automatic model detection • Bug reporting feature added to Settings > About for easy feedback submission to GitHub IMPROVEMENTS • Cleaner, more professional tutorial and settings language without overwhelming technical platform lists • Menu bar now shows meetings within 15-minute window instead of 1 hour for more relevant notifications • Recording filenames use meeting title captured at recording start time to handle overtime meetings correctly • Enhanced purchase messaging emphasizes supporting app development • Automatic model discovery from AI services with intelligent caching • Enhanced Settings UI with dropdown model pickers, Discover buttons, and vision model indicators BUG FIXES • Restored Zoom detection after multi-platform refactor regression Questions or feedback? Use the new Report Bug feature in Settings > About!

Expanded Meeting Platform Support MeetingMind now automatically detects and records meetings across Microsoft Teams, Google Meet, and Webex—in addition to Zoom. Your meetings are captured seamlessly, no matter which platform you use. More Recording Time The free tier now includes 60 minutes of recording per day, doubled from 30 minutes. Enjoy more flexibility to capture your important conversations. Improved User Experience - What's New window now properly adapts to light and dark mode - Cleaner version history display with smooth animations - Simplified interface automatically manages update notifications Performance Enhancements - Improved audio system responsiveness - Fixed performance warnings for smoother operation - Better thread management for more reliable recordings

Various improvements to recording and calendar support.