Recorder Changelog

​

🚀 New Features

• [Desktop] OpenAI Speech-to-Text: Use OpenAI for live transcription and for retranscribing saved recordings, with model selection.
• [Desktop] Video Meeting Recording Controls: Settings include controls for video + audio or audio-only capture, plus clearer availability and setup status.
• [iOS] OpenAI Speech-to-Text: OpenAI live transcription and retranscription on iPhone, with model selection.
• [iOS] On-Device Transcription on iPhone: Parakeet on-device transcription with live + post-recording modes and built-in model download/management.
• [Shared] Cloud Model Picker: Choose between cloud AI model options for supported providers directly in Settings without re-entering your API key.
• [Shared] Retranscription Progress Indicator: Live progress updates while regenerating a transcript.
• [Shared] AI Quality Indicator: Chat and summary screens show an AI quality badge with model details and a shortcut to Settings.

​

✨ Improvements

• [Desktop] Clearer Video Meeting Recording Language: Simpler, friendlier wording in settings and status messages.
• [Shared] Better Transcription Setup Guidance: Setup screens better show when OpenAI or Parakeet is ready and what’s missing.
• [Desktop] Easier Parakeet Setup: Clearer model status, download progress, upgrade prompts, supported-model guidance.
• [Shared] Smarter Summary Progress Feedback: Clearer progress and completion states while summaries generate.
• [Desktop] Improved Audio Device Handling: Default mics/speakers marked clearly; system device changes followed reliably during recording.
• [Desktop] More Reliable Dictation Behavior: Dictation waits briefly for trailing speech before stopping, preserving last words.
• [Desktop] Improved macOS Permission Flow: Screen recording and camera permission messaging is clearer and refreshes reliably after grant/deny.
• [Desktop] Better macOS Update Guidance: Warns when the app runs from a location where automatic updates won’t work (e.g., outside Applications).
• [iOS/Android] Smoother Mobile Startup and Navigation: More reliable mobile startup, clearer loading/error states, more responsive onboarding and recording.
• [iOS] Faster Mobile Recording Response: Mobile recording feels quicker and less likely to stall on start/stop.
• [iOS] More Reliable Saved Recordings: iPhone recordings preserved more safely, reducing accidental loss from temporary storage.
• [Desktop] Improved Local AI Output Formatting: Local AI responses preserve spacing and formatting (lists, indentation) more accurately.
• [Desktop] Clearer Local AI Error Messages: Local AI requests that are too large get clearer fix guidance.
• [Desktop] Better Long Transcript Progress Updates: Step-by-step progress on long retranscriptions instead of appearing stuck.
• [Desktop] Improved Transcript Panel Layout: Transcript controls behave better on small desktop window sizes.
• [Shared] Updated Transcription Quality Ratings: Clearer 7-level scale for comparing privacy and accuracy tradeoffs between providers.

​

🐛 Bug Fixes

• [Shared] More Reliable Final Transcript Capture: Stopping a recording or dictation is less likely to cut off the last few spoken words.
• [Shared] Better Recovery from Cloud Transcription Disconnects: Cloud transcription recovers more reliably from connection issues and preserves more transcript content.
• [Shared] Fixed Empty Summary Results: Failed summary generation now reports the problem instead of finishing silently with nothing saved.
• [iOS] Fixed White Screen and Startup Crashes: Mobile startup more stable; fixed blank screens, init failures, render crashes.
• [iOS] Fixed Recording State Issues: App stays in sync with actual recording state more reliably during slow starts or failed stops.
• [iOS] Fixed Loss of Recordings: Existing iPhone recordings moved into durable storage more reliably.
• [iOS] Fixed Parakeet Final Transcript Loss: Stopping a Parakeet recording flushes remaining audio more safely so final speech isn’t lost.
• [Desktop] Fixed Parakeet Retranscription for Saved Recordings: Regenerating transcripts now works correctly for Parakeet recordings, including long ones that previously could fail or crash.
• [Desktop] Fixed Speaker Labels in Retranscription: Retranscribed recordings with speaker detection keep more of the transcript and apply speaker labels more accurately during fast speaker switches.
• [Desktop] Fixed macOS Camera Permission Issues: Webcam overlay and camera access on macOS request and report permissions more reliably.
• [Desktop] Fixed Auto-Update Problems on macOS: App avoids install attempts from unsupported locations and shows clearer guidance instead of failing silently.
• [Shared] Fixed Authentication Refresh Glitches: Intermittent sign-in failures from overlapping refresh attempts are less likely.
• [Desktop] Fixed Local AI Hangs and Broken Characters: Local AI generation more stable; less likely to freeze or produce malformed text.
• [Desktop] Fixed Missing Transcript Panel Controls: Close and copy buttons in the transcript view remain visible more reliably.
• [Desktop] Fixed Popovers Near Screen Edges: Quality and settings popovers display correctly on small or narrow windows.
• [Desktop] Fixed Floating Bar After Switching from Notch: Switching window style from notch to floating no longer leaves the bar clipped — geometry resets fully without an app restart.
• [Desktop] Fixed Notch After Onboarding: Finishing onboarding with the notch style no longer shows a squashed, upside-down floating bar — the main window switches to the notch UI correctly.
• [Desktop] Fixed BuildBetter Transcription Hang: A stalled credits check or WebSocket handshake no longer blocks live transcription forever — the app gives up after 45 seconds with a clear “session start timed out” error.
• [Desktop] Fixed Notarization for Bundled Frameworks: Release builds with the video meeting recording SDK now sign every nested binary individually with a secure timestamp, so Apple notarization no longer rejects inner GStreamer dylibs.

​

🚀 New Features

• [Desktop] Retranscription Progress Updates: Live status updates in the transcript pane — loading audio, uploading, transcribing, saving.
• [Desktop] Cloud Model Picker: Choose between available cloud AI models in Settings → AI without re-entering API keys.
• [Shared] AI Quality Indicators: Chat, summaries, and transcripts show clearer quality and provider details — which model or transcription engine was used and what to expect.
• [iOS] Expanded iOS AI Options: Choose on-device AI, BuildBetter, or your own OpenAI / Anthropic / Groq keys for summaries and chat on iPhone/iPad.
• [iOS] iOS API Key Management: Add, review, and remove supported AI API keys directly in Settings with secure storage.
• [iOS] BuildBetter Transcription on iOS: iOS supports BuildBetter as a transcription option, including live transcription and cloud retranscription.
• [Shared] Back-to-Back Recording Preferences: Choose how overlapping meetings are handled — prompt, auto-split, or keep as one.
• [Desktop] Recording Split, Merge, and Archive Actions: Supported recordings include user-facing options to split, merge, or archive from Recording Detail.
• [Desktop] Archived Recordings: Archive instead of delete; hidden from the main list, restorable later.
• [Desktop] Clear Chat: Recording chat threads can be cleared from the chat header.
• [Desktop] macOS Meeting Notifications: Notified when a new meeting starts while another recording is in progress.
• [Desktop] Speaker Identification Banner: Transcripts show a clearer in-app prompt when speaker identification is off, with a one-click enable.
• [Desktop] Setup Wizard Feedback Button: Send feedback directly from the setup flow.
• [Shared] Post-Call Rating Controls: Disable rating prompts from Settings or pick “Don’t ask me again” in the rating dialog.
• [Desktop] macOS Screen Recording: Record a full display or single window on macOS, with optional mic audio, system audio, and webcam overlay.
• [Desktop] Screen Recording Setup Window: Dedicated setup for sources, audio options, and camera placement.
• [Desktop] Screen Recording Library: Screen recordings have their own library view, separate from audio.
• [Desktop] Goal-Based Onboarding: Onboarding adapts to whether you want meeting recording, voice capture, and/or screen recording.

​

✨ Improvements

• [Desktop] Clearer Retranscription Warnings: Regenerate-transcript dialog explains speaker-identification tradeoffs.
• [Desktop] Better Speaker Paragraphs: Retranscribed transcripts grouped into cleaner speaker paragraphs.
• [Desktop] More Accurate Speaker Assignment: Better labeling around short interjections and overlapping speech.
• [Desktop] Improved Long Recording Handling: Long retranscriptions process more smoothly with better progress updates and fewer backend limits.
• [Desktop] Faster Parakeet on Apple Silicon: Local Parakeet retranscription uses Apple hardware acceleration when available.
• [Desktop] More Responsive Transcript Panel: Controls stay visible at narrower window sizes.
• [Shared] Updated Model Names and Defaults: Names, descriptions, and defaults refreshed to reflect current options and quality.
• [Desktop] Smoother Settings Experience: AI model lists load smoothly and preserve existing API key details when you switch models.
• [Desktop] Better Transcript Quality Details: Transcript status shows which provider, model, and speaker-labeling method were used, with safe fallbacks for older recordings.
• [iOS] Streaming AI Responses on iOS: Cloud AI responses stream live for more responsive chat.
• [Shared] Better BuildBetter Reliability: BuildBetter transcription more resilient to temporary disconnects; recovers automatically in more cases.
• [Shared] Credit Exhaustion Alerts: Clearer warnings when BuildBetter credits run out, with a quick path to switch settings.
• [Shared] Automatic BuildBetter Defaults: Eligible workspace users get BuildBetter selected automatically for transcription and snippet cleaning, preserving manual choices.
• [iOS/Android] Improved Mobile Experience: Classic and Retro themes, theme switcher, better playback controls, mobile retranscription, configurable mic sample rate.
• [Desktop] Smarter Recording Actions: Split/merge/archive include clearer confirmations and hide when they don’t apply.
• [Desktop] Shorter “Catch Me Up” Recaps: Concise three-bullet summary.
• [Shared] Improved Onboarding Messaging: Setup explains AI quality, speed, download size, permissions, and speaker-identification tradeoffs more clearly.
• [Desktop] Visible Diarization Download Size: Speaker identification download size shown upfront during setup.
• [Shared] Less Disruptive Rating Prompts: Post-call rating prompts appear after recording review, not immediately at end.
• [iOS] Refreshed iOS Onboarding: Cleaner, more native look with clearer setup steps.
• [iOS/Android] Classic Theme by Default: New mobile installs start with the Classic theme.
• [Desktop] Better Screen Recording Workflow: More consistent behavior from setup through final processing; completed videos used reliably for playback, upload, sharing.
• [Desktop] Screen Recordings Get Post-Processing: Same titles and summaries treatment as audio recordings when transcripts are available.
• [Desktop] Improved Folder Management: Clearer folder handling with a visible main Recordings folder, better filtering, and rename/move support.
• [Desktop] More Reliable Dashboard Navigation: Opening directly to areas like Recordings works more consistently.

​

🐛 Bug Fixes

• [Desktop] Full Transcript Preservation: Retranscription with speaker identification no longer drops text from short segments.
• [Desktop] Safer Diarization Fallbacks: If speaker identification fails, the app keeps the full transcript instead of returning partial.
• [Desktop] Fixed Speaker Label Collapsing: Retranscribed conversations no longer collapse into a single “Speaker 0” block when speaker separation is available.
• [Desktop] Parakeet Retranscription Restored: Parakeet recordings can be retranscribed correctly again.
• [Desktop] Long Parakeet Crash Fix: Long local Parakeet retranscriptions no longer crash the desktop app.
• [Desktop] Transcript Controls Visible Again: Close and Copy buttons in the transcript panel no longer disappear at medium window sizes.
• [Desktop] Quality Popover Positioning: Popovers stay readable instead of being clipped at window edges.
• [Desktop] Older Recording Metadata Fallbacks: Older recordings without saved provider details show safer fallback info instead of misleading labels.
• [iOS] iOS Theme Onboarding Fix: Onboarding always uses the intended appearance, even if a different theme was previously saved.
• [iOS] Version Display Fix: Version and build info correctly match the current release.
• [Shared] BuildBetter Speaker Numbering Fix: Retranscribed speakers now start at Speaker 1 for consistency.
• [Desktop] Settings Shortcut Fix: “Open Settings” from credit warnings works more reliably, even from the recorder window.
• [iOS/Android] Mobile Reliability Fixes: Better onboarding layout on modern iPhones, more reliable feedback submission, improved long-recording stability, better transcript finalization on end.
• [Shared] BuildBetter Login Fallback: Selecting BuildBetter without the required login or workspace falls back safely instead of failing silently.
• [Desktop] Auto-Split Timing Fix: Back-to-back meeting auto-splitting happens at the correct time.
• [Desktop] Invalid Split/Merge Protection: Unsupported split/merge actions blocked safely to prevent broken results.
• [Shared] Token Refresh Reliability: Authentication refreshes more reliably before expiry, reducing intermittent failures.
• [Desktop] BuildBetter Retranscription Fix: Saved recordings use BuildBetter correctly for retranscription instead of falling back to the wrong backend.
• [Shared] BuildBetter Transcript Fidelity: Retranscriptions better preserve speaker labels, timing, and confidence.
• [iOS] Calendar Loading Feedback on iOS: “Show calendars” setting gives clearer loading, retry, and empty-state feedback.
• [Shared] Deepgram Language Fixes: Selected languages and auto-detect work correctly across live and retranscription, improving multilingual accuracy.
• [Desktop] Faster “No Transcript Found” Feedback: Recordings that never produce a transcript surface the result sooner.
• [Desktop] Improved Screen Recording Stability: Crashes, preview issues, shutdown reliability, and recording state sync fixed on macOS.
• [Desktop] Smoother Live Cursor Preview: Mouse cursor smoother during live screen recording previews.
• [Desktop] Camera Overlay Position Fix: Webcam overlays stay correctly positioned in the recorded video.
• [Desktop] Long Screen Recording Upload Fix: Screen recordings with webcam overlay less likely to miss auto-upload after long export times.
• [Shared] Key Persistence Fixes: Sign-in tokens and saved cloud API keys recover more reliably when secure storage is flaky.
• [Desktop] Folder Operation Recovery: Folder moves, renames, and creation behave more reliably if something fails partway through.

​

🚀 New Features

• [Desktop] Screen Recording on macOS: Record full screen or a single window, with optional mic audio, system audio, and webcam overlay.
• [Desktop] Screen Recording Setup Window: Choose what to record, preview sources live, adjust audio options, and position your camera overlay before you start.
• [Desktop] Dedicated Screen Library: Screen recordings have their own Screen section.
• [Desktop] Live Camera Preview: Floating live camera preview during capture when recording with webcam.
• [Desktop] Goal-Based Onboarding: Setup adapts to your goals — meetings, talk-to-text snippets, or screen recording.
• [Desktop] Recording Folders: Create folders, set a default folder, and move recordings between folders.
• [Desktop] Groq AI Provider Support: Groq available as a cloud AI option for chat, summaries, dictation cleanup, and transcription.
• [Desktop] Dictation Finalize Mode: Pick whether dictation pastes raw transcript immediately or waits for a cleaned-up final version.
• [Desktop] Meeting Note Prompt: Prompt to start taking notes when a meeting starts.
• [Desktop] Meeting Trigger Details: Recordings show why they started (meeting auto-start, dictation trigger).
• [Desktop] Mirror Camera Option: Mirror the webcam overlay in screen recordings.
• [iOS] Classic and Retro themes: iOS theme options with a new theme switcher in Settings.
• [iOS] Dedicated playback controls: Scrub, skip, and playback-speed controls on iOS.
• [iOS] Retranscription on iOS: Existing recordings can be retranscribed from the mobile app.
• [iOS] Configurable mic sample rate: Pick the sample rate used when recording on iOS.

​

✨ Improvements

• [Desktop] Smarter Screen Recording Processing: Auto titles and summaries when usable audio is available.
• [Desktop] Better Screen Recording Playback: Detail pages prioritize the finished video, support inline playback, and still expose the raw camera clip.
• [Desktop] Improved Screen Recording Flow: Starting and stopping returns you to the right screen more consistently.
• [Desktop] More Reliable Dashboard Navigation: Opening directly to tabs like Recordings is more dependable.
• [Shared] Clearer Onboarding: Refreshed across desktop and iOS — clearer wording, better permission guidance, cleaner layouts, step-by-step progress indicator.
• [Desktop] Simpler Permission Requests: Permission prompts explain required vs. optional, especially for microphone, calendar, camera, and screen recording.
• [Desktop] Hosted Workflow Setup Improvements: BuildBetter workspace setup loads earlier and remembers your selected workspace.
• [Desktop] Improved Dictation Speed and Reliability: Finalization faster, more consistent, and better at preserving cleaned text, source app, and tone history.
• [Desktop] Richer Transcript Formatting: Better preservation of speaker labels, punctuation, timing, confidence, and other metadata.
• [Desktop] More Reliable Transcript Saving: Live transcripts continuously saved during recording, improving summary accuracy.
• [Desktop] Better Summary Accuracy: Auto-summaries use the most complete transcript available.
• [Desktop] Improved “Catch Me Up”: Catch-up summaries focus on the last five minutes and return three short bullets.
• [Desktop] Faster AI Settings Refresh: Changing cloud AI API keys refreshes available models quickly and reliably.
• [Desktop] Improved Feedback Reporting: Feedback and post-call reports include better diagnostics for faster resolution.
• [Desktop] Window Style Preference Persistence: Desktop window style restores reliably between launches.
• [Desktop] Smoother Desktop Startup: More stable startup, fewer visual glitches, more consistent window behavior.
• [Desktop] Better Audio Stability During Calls: Microphone and system audio handling on macOS more stable during device changes.
• [Desktop] Improved Meeting Detection: Fewer false triggers from background activity; more accurate start/stop detection.
• [Desktop] Streamlined Recordings Experience: Recent recordings and navigation more consistent.
• [Desktop] Safer Storage Management: Settings clearer about where recordings save; storage usage reporting more accurate.
• [iOS] Cleaner iOS Default Theme: New iOS installs default to Classic.
• [iOS] Major mobile UI overhaul: Redesigned recording, playback, and settings.
• [iOS] Settings rewrite: Refreshed layout with clearer sections, better copy, new controls.
• [iOS] Recording detail polish: Inline playback, upgraded transcript viewing, easier sharing.
• [iOS] Updated on-device intelligence library: Better accuracy and stability.
• [iOS] Richer transcript view: Better handling of speaker labels, punctuation, and long-running sessions.

​

🐛 Bug Fixes

• [Desktop] Faster “No Transcript Found” Handling: Recordings that never receive a transcript fail sooner.
• [Desktop] More Reliable Screen Recording Startup: Fixed cases where screen recording appeared stuck, failed to start, or prompted for the wrong permissions.
• [Desktop] Improved Screen Recording Stability: Crashes, corrupted previews, startup failures, unreliable stopping/finalization all fixed.
• [Desktop] Camera Overlay Positioning Fixed: Webcam picture-in-picture overlay on macOS stays in place.
• [Desktop] Smoother Cursor Capture: Mouse movement smoother in macOS screen recordings.
• [Desktop] More Reliable Camera Overlay Recording: Fixed cases where the live camera preview or final exported video could miss the camera feed.
• [Desktop] Correct Camera Overlay Layering: Camera overlay properly appears on top of the screen recording.
• [Desktop] Better Long Recording Finalization: Long screen recordings with camera overlay less likely to fail upload or finish processing late.
• [Desktop] System Audio Permission Fix: Turning off Include system audio properly skips capture and avoids unnecessary prompts.
• [Desktop] Unsupported Device Handling: Devices/macOS versions that don’t support screen recording are identified correctly.
• [Desktop] Screen Tab Folder View Fix: Screen recordings remain visible when browsing within a folder.
• [Desktop] Recording Routing Fixes: Fixed cases where recording windows or shortcuts opened the wrong screen.
• [Desktop] Summary Refresh Fixes: Auto-summaries appear reliably in recording details without manual refresh.
• [Desktop] Transcript Save/Load Fixes: Fixed missing, misplaced, incomplete, or wrongly attached transcripts.
• [Desktop] Live Transcript Reliability: Fixed cases where live recordings could end with empty transcripts despite captured speech.
• [Desktop] Apple Speech Stability Fixes: On-device transcription recovers better if it stalls during recording.
• [Desktop] Sign-In and API Key Persistence: Tokens and saved API keys recover reliably when macOS Keychain is flaky.
• [Desktop] Window Dragging Restored: Dashboard title bar can be dragged again without interfering with buttons, links, or search.
• [Desktop] Window Style Preference Fix: Window style no longer unexpectedly switches after upgrading.
• [Desktop] Folder Safety Improvements: Invalid folder names/IDs blocked; folder moves safer mid-save.
• [Desktop] Reduced False Meeting Starts: Webex background activity less likely to trigger accidental prompts.
• [iOS] iOS Onboarding Layout Fixes: Spacing and safe-area issues fixed on iPhones with notches or home indicators.
• [iOS] More reliable feedback submission: iOS feedback batched and submitted more reliably.
• [iOS] Mobile capture memory fix: Memory allocation bug in the mobile capture layer fixed.
• [iOS] Live transcript finalization on iOS: iOS live transcripts preserve captured text more reliably on end.
• [Desktop] Dictation Cleanup Fix: Finalized dictation no longer includes internal system messages in pasted text.

​

🚀 New Features

• [Desktop] Recording folders in the sidebar: Create folders from the sidebar, with a visible Recordings folder for everything in the main library, default-folder selection, and folder filtering.
• [Desktop] Notion export beta: Connect Recorder to Notion, choose a default destination, push notes/summaries/transcripts from Recording Detail.
• [Desktop] Groq cloud support: Groq as a cloud provider for AI chat, summarization, snippet cleanup, live transcription, and retranscription.
• [Desktop] Meeting-aware recording flow: Auto-started recordings link to the current calendar event; upcoming meetings open their linked recordings; Recorder can prompt for a fresh recording when meetings overlap.
• [Desktop] Meeting relink controls: Recording Detail can suggest nearby meetings and lets you relink/unlink mis-matched ones.
• [Desktop] Take-notes meeting prompt: Native prompt when a detected meeting starts.
• [Desktop] Limited onboarding mode: Setup continues in reduced-capability mode when permissions or models are still missing.
• [Desktop] Custom snippet cleaner instructions: Give the dictation cleaner extra guidance for your style.
• [Desktop] Flexible storage controls: Choose recording location, set default folders, optionally compress finished local audio.

​

✨ Improvements

• [Desktop] Folder management now lives where you use it: Folder creation pulled out of Settings and into the sidebar.
• [Desktop] Better folder settings: Rename, change save location, optionally move existing recordings.
• [Desktop] Faster move-to-folder flow: Show in Finder and Move to folder combined into one menu.
• [Desktop] Better folder filtering and counts: Browse and search custom folders or main Recordings more reliably.
• [Desktop] More flexible Notion defaults: Remember preferred connection, parent page, and exported sections.
• [Desktop] Clearer integrations labeling: BuildBetter and Notion connections live under Integrations.
• [Desktop] Richer playback controls: Scrub, skip back/forward, playback speed changes.
• [Desktop] Smarter transcript recovery: AI chat and transcript actions fall back gracefully when transcript files are missing.
• [Desktop] More reliable uploads and feedback flows: Common media types preserved cleanly; feedback dialogs remember your email.
• [Desktop] Clearer AI assistant readiness: Explains why AI chat is unavailable and points you to the right setup step.
• [Desktop] More flexible BuildBetter environment routing: Desktop auth, uploads, and integrations behave consistently across prod, staging, and local environments.
• [Desktop] Better recording provenance details: Developer Mode shows why a recording or snippet was created.
• [Desktop] More control over meeting prompts: Manage take-notes pop-up from Calendar settings; preview in Developer Mode.
• [Desktop] Clearer update guidance: Update failures from disk-image or translocated paths explained.
• [Desktop] Clearer transcription language guidance: Setup flags English-only Whisper model mismatches and suggests multilingual alternatives.
• [Desktop] Faster AI settings refresh after key changes: Saving/removing a cloud API key refreshes model availability immediately.
• [Desktop] Better summary automation organization: Auto-title and auto-summary controls alongside summary templates.
• [Desktop] Clearer snippet browsing: Long snippet titles and previews wrap cleanly.
• [Desktop] Cleaner dictation output: Cleanup preserves your wording, removes repeated passages, adds natural paragraph breaks.
• [Desktop] Smoother live transcript updates: Long recordings stream transcript updates with less lag and UI churn.
• [Shared] More focused catch-up summaries: Catch-me-up returns three short bullets based on the last five minutes.
• [Desktop] More resilient dictation overlays: Dictation Bar stays compact until speech arrives; Notch recovers reliably if cleanup or stop events delay.
• [Desktop] Clearer auto-paste guidance: Errors explain disk-image, translocated path, or local dev build issues.
• [Desktop] More helpful empty states: Home can start a recording directly when your library is empty.
• [Desktop] More reliable macOS packaging: Cleaner updater archives and more reliable DMGs on newer macOS.
• [Desktop] Safer updater artifacts: Versioned updater bundles reduce cache collisions between releases.
• [Shared] Consistent app naming across releases: macOS keeps BuildBetter.ai Recorder, iOS keeps BB Recorder.

​

🐛 Bug Fixes

• [Desktop] Cleaner recording shutdown timing: Waits for final audio and transcript output before announcing stop.
• [Desktop] Safer folder moves and renames: Rolls back safely if a file move fails partway through.
• [Desktop] More reliable folder creation: Folder creation, library counts, and default-folder handling more consistent.
• [Desktop] More reliable Notion environment handling: Notion export follows the BuildBetter environment of your sign-in.
• [Desktop] Fewer updater install failures: Update archives strip problematic macOS metadata files.
• [Desktop] More robust build environment checks: Local build scripts catch missing Xcode toolchain earlier.
• [Desktop] Safer transcription language validation: Unsupported codes rejected earlier with clearer errors.
• [Desktop] Fewer accidental dictation hotkey triggers: Fn, Globe, and other modifier hotkeys less likely to start/stop dictation unexpectedly.
• [Desktop] Fewer false-positive meeting auto-starts: Webex background activity less likely to mimic an active meeting.
• [Desktop] Safer recording state transitions: Guards against duplicate starts; keeps more trailing transcript text before finalizing.
• [Desktop] Better recovery when capture ends unexpectedly: Stops and cleans up reliably if audio sources disconnect or transcript streams stall.
• [Desktop] More reliable floating window behavior: Manually opened windows stay visible until dismissed.

​

🚀 New Features

• Snippets mode for dictation: Capture quick dictation clips, then browse and copy them from the Snippets view.
• New Dictation Bar and notch controls: Faster start/stop with dedicated dictation surfaces.
• Dictation & Snippets settings page: Configure hotkeys, auto-paste, always-visible bar, and snippet display preferences in one place.
• Template-based auto-summarize defaults: Choose built-in or custom templates as the default auto-summary format.
• Snippet cleaner provider/model selection: Pick how cleanup runs before paste — local models, Apple Foundation, OpenAI, Anthropic, BuildBetter, or Ollama. Includes GPT-5 Nano.
• Parakeet transcription option: New transcription engine with guided model management.
• BuildBetter cloud support: Use BuildBetter-hosted AI for transcription and chat when signed in.
• Enhanced Notes tab: Dedicated enhanced-notes view in recording detail.
• Built-in diagnostics export: Generate diagnostics and open logs from settings.

​

✨ Improvements

• More reliable meeting auto-detection: Better signal handling for start/stop.
• Smoother dictation hotkey behavior: More consistent hold, tap, and double-tap flows.
• Cleaner transcript finalization: Dictation waits more intelligently for trailing words before cleanup and paste.
• Faster dictation stop-to-paste flow: Reduced stop tail and settle windows in Notch and Dictation Bar modes.
• More stable notch transcript layout: Less resize jitter and visual flicker while dictating.
• Clearer setup guidance: AI setup and settings explain readiness, permissions, and model choices.
• Consistent templates across settings and recordings: Same catalog powers template settings, default summary, and recording summaries.
• Better update visibility: Install/restart actions and update state clearer.
• Faster recovery under load: Transcription streams handle temporary queue pressure better.

​

🐛 Bug Fixes

• Reduced repeated macOS permission prompts: Accessibility prompts less repetitive.
• Action items render correctly as lists: Bulleted action items format as expected.
• Safer stop handling: Stopping when nothing is recording no longer causes confusing states.
• Cleaner cleanup fallback behavior: Dictation cleanup rejects meta/model chatter and falls back to cleaned raw text.
• No duplicate dictation surfaces in notch mode: Prevents overlapping dictation UI.
• Clearer sign-in/environment errors: Auth errors easier to read when environment settings don’t match.
• Meeting auto-start defaults tuned for reliability: Existing installs use more reliable defaults.
• Safe fallback when deleting a default template: Switches safely to Quick Summary if a default template is removed.
• Launch-at-login toggle stability: Setting persists correctly.
• No stale restart prompts: Restart/update prompts don’t appear when already up to date.
• Calendar deselect-all now sticks: Empty selections preserved.
• Auto-stop no longer ends manual recordings/snippets: Auto-stop applies only where expected.

​

🚀 New Features

• Comprehensive Transcription Readiness Check: New get_transcription_readiness command shows exactly what’s working and what needs setup — checks macOS Speech Recognition (availability + permission), Whisper (installed + active model), and Deepgram (API key). Returns a clear isReady boolean with specific actionNeeded guidance.
• Parakeet Transcription Backend (feature-gated): NVIDIA Parakeet TDT 0.6B via sherpa-onnx — no Python dependency, ~1.93% word error rate, single-model caching (~3 GB per model).
• Download Progress Bars: Model downloads in onboarding show real progress bars with percentage.

​

✨ Improvements

• Transcription Status Cards: AI Settings and Setup Wizard show clear status cards — green for ready, amber for needs-setup, gray for unavailable.
• Enhanced Debug Panel: “Check Permissions” tests all 6 TCC permissions (Microphone, Screen Recording, Accessibility, Automation/AppleEvents, Calendar, Speech Recognition) and shows transcription readiness.
• Reset Permissions Updated: Resets all 6 TCC permissions including Calendar, AppleEvents, SpeechRecognition.
• Prominent Backend Logging: Clear ======== banner logs when transcription starts showing which backend is in use.

​

🐛 Bug Fixes

• Fixed Apple Speech duplicate transcriptions: Switched from string prefix matching (lastEmittedText) to segment-based tracking (lastEmittedSegmentCount) — only new segments emitted.
• Fixed memory leak / infinite tick logs: WebSocket transcription server loop now exits properly via explicit None handling in tokio::select!.
• Fixed CSS @keyframes nesting: Resolved CSS compilation warnings in classic theme.

​

🔧 Technical

• New TranscriptionReadiness and TranscriptionOptionStatus types in commands/apple_intelligence.rs.
• Segment-based transcription tracking in AppleIntelligence.swift.
• Proper tokio::select! None handling pattern in transcription/server.rs.
• Parakeet module with sherpa-rs TransducerRecognizer integration.

​

🚀 New Features

• Ollama Integration: Local Ollama models for AI Assistant and Summarization. Auto-detects server at localhost:11434, lists all available Ollama models in AI engine dropdowns, available in the onboarding wizard, streaming generation with fallback support.
• Custom Whisper Model Import: Import your own .bin Whisper models. Copy to app folder or reference external paths. Imported models appear in the dropdown with “(Custom)” badge. GGML files validated before import.
• NotchBar External Monitor Support: NotchBar always stays centered on the MacBook’s built-in display. Screen observer monitors display configuration changes, uses CGDisplay.is_builtin(), auto-repositions when external monitors connect/disconnect.
• Speaker Diarization (WIP): Infrastructure for speaker identification in transcripts.

​

✨ Improvements

• Model Download Validation: Whisper downloads use a temp .download file first, validate file size, auto-detect and re-download corrupted/incomplete files.
• Better Transcription Error Messages: Errors include file size for debugging; warns if file is suspiciously small; instructs to delete and re-download.
• Distil-Whisper Model Priority: Distil models checked first in discovery (ggml-distil-small.en.bin, ggml-medium-32-2.en.bin, ggml-distil-large-v3.bin).
• Apple Intelligence Detection: Clearer messaging — macOS version detection (consumer 15.1+ vs Foundation Models API 26+), “How to verify” expandable guide, “Open System Settings” button.
• Upload Button Auth Check: Shows “Login to Upload” when not authenticated; toast directs to Settings.
• Auth Environment Validation: Clear error when staging/production are mismatched: “You are logged into staging but trying to use production”.
• Live Changelog Updates: In-app changelog fetches from server; always up to date regardless of installed app version; falls back to embedded version when offline.

​

🐛 Bug Fixes

• Auto-stop now works when meeting ends: Calendar mode requires activity signals (mic/camera/network) for high confidence — when you leave a meeting, signals stop, confidence drops to 0.20, and auto-stop triggers. Confidence scales by participation: Calendar + mic + camera = 0.99, Calendar only (no signals) = 0.20.
• Calendar selection now syncs: Toggles emit events; UpcomingView updates immediately without restart.
• Fixed app crash on startup: Screen observer uses correct async runtime (tauri::async_runtime::spawn).
• Whisper model not found: Model discovery includes all model types with detailed search-path logging.
• Retro NotchBar stop button overlap: Removed LED indicator and adjusted position.

​

🚀 New Features

• Custom Whisper Model Import: Import your own .bin Whisper models. Copy to app folder or reference external paths. Imported models appear in the dropdown with “(Custom)” badge.
• Ollama Integration: Ollama models for AI Assistant and Summarization. Auto-detects at localhost:11434, lists available models in dropdowns, available in onboarding wizard, streaming with fallback.

​

🐛 Bug Fixes

• Fix Apple Intelligence not detected on macOS 26 (dylib bundled in Frameworks/ but code looked in Resources/).
• Fix Whisper model import file picker not opening (AppleScript command fix).
• Fix microphone staying active after recording ends on macOS (CPAL stream explicitly paused before drop).
• Fix Whisper transcript overwriting single line instead of growing (is_final: false → true).
• Fix stop button not responding on recordings page (isStopping state reset in finally block).
• Fix “Show Recorder” tray menu not bringing window to front on macOS (added activateIgnoringOtherApps).

​

✨ Improvements

• Apple Intelligence Detection: macOS version detection between consumer features (15.1+) and Foundation Models API (26+); “How to verify Apple Intelligence” expandable guide; “Open System Settings” button; improved error messaging when API unavailable.

​

🚀 New Features

• Hide Dock Icon: New “Menu Bar Only” mode hides the app from Dock and Cmd+Tab — accessible only via the system tray (macOS).
• Retranscribe Recordings: Re-transcribe existing recordings using any backend (Whisper, Deepgram, Apple Speech).
• Calendar Selection: Choose which calendars appear in Upcoming Meetings with checkboxes.
• [iOS] In-App Feedback: Submit feedback with star ratings directly from the mobile app.

​

🐛 Bug Fixes

• Fix calendar permission dialog not appearing on macOS 14+ (requestFullAccessToEventsWithCompletion:).
• Fix “Invalid Date” display in calendar events (serde camelCase serialization).
• Fix calendar settings button navigating to the wrong tab.
• Fix delete recording button not responding to clicks.

​

✨ Improvements

• Calendar Settings shows selected calendar count and supports select/deselect all.
• System tray menu includes “Toggle Dock Icon” option.
• Added Checkbox UI component for better form controls.

​

📱 Mobile App

• New in-app feedback form with star rating and Slack/API submission.
• Retranscribe recordings via RefreshCw button on detail screen.
• CollapsibleBottomBar styling improvements.
• Recordings and Settings screen refinements.

​

🚀 New Features

• Theme system with Classic and Retro styles.
• Templates tab in Settings with custom summary templates.
• Changelog UI with version history and provider logos.
• Deepgram cloud transcription enhancements.

​

🐛 Bug Fixes

• Fix theme switching — remove global retro styles from index.css.
• Fix speaker diarization display; add dev mode option for rating popup.
• Fix Deepgram interim transcript handling and consolidate transcript views.
• Fix system audio banner not clearing when audio detected.
• Fix NotchBar expand/move issues on button clicks.
• Fix Deepgram duplicate transcriptions by mixing audio before sending.
• Remove broken latency tracking from live transcription.

​

✨ Improvements

• Desktop components theme-aware with CSS variables.
• Reload LLM and transcription backends when live recording starts.
• Simplify NotchBar by removing expand/quick-actions view.
• Color-code transcript bubbles by speaker.
• Redesign templates UI and add template grid to summary view.

​

🚀 New Features

• Theme selection (Classic / Retro) added to the onboarding wizard.
• Classic theme: clean modern look with purple accents.
• Retro theme: terminal-inspired with amber/green and VT323 font.

​

🐛 Bug Fixes

• Fix NotchBar buttons staying in active state after click.
• Fix default theme now correctly set to Classic.
• Fix gray-on-gray text in template section badges.
• Fix onboarding showing retro despite Classic being selected.
• Fix CSS @import order warnings for Google Fonts.

​

✨ Improvements

• Streamlined onboarding (removed skip button).
• Theme-aware components using CSS variables.
• Centralized theme defaults to classic.

​

🚀 New Features

• Apple Speech improvements with better transcription quality.
• Copy transcript to clipboard functionality.
• LLM token cleanup for better context management.

​

✨ Improvements

• Reload transcription backend when settings change.
• Better speaker color coding in transcripts.

​

🚀 New Features

• Auto-Titles: AI automatically generates descriptive titles from transcripts.
• Markdown Notes: Edit/preview toggle with full GitHub Flavored Markdown support.
• Cloud API Keys (BYOK): Bring your own keys for OpenAI, Anthropic, and Deepgram.
• Custom Summary Templates: Create reusable meeting note formats.
• NotchBar shows all 4 quick actions in expanded view.

​

🔧 Technical

• Cloud passthrough commands for OpenAI / Anthropic / Deepgram.
• LLM router supports multiple backends.
• Cleaned up unused legacy components.
• Template storage via plugin-store.

​

✨ Improvements

• Prominent Refresh button on AI Engines settings.
• Toast notifications for model refresh.
• Auto-fallback to Apple Intelligence when active model deleted.
• Refresh button for Apple Intelligence status.
• Apple Intelligence default in onboarding when available.
• Apple Intelligence setup prompt added to the onboarding wizard.
• Apple Intelligence setup guide in AI Settings.

​

🐛 Bug Fixes

• Fix release build hang issue.

​

🔧 Technical

• Bump version and update Apple Intelligence dylib.

​

🚀 New Features

• AI Engines System: Separate configuration for different AI providers.
• LLM Router: Intelligent routing between local and cloud AI models.
• Advanced Summarization UI: Dropdown menu for summary options.

​

🔧 Backend

• Catch-me-up feature (summarize last N minutes).
• Progressive summarization for long transcripts (map-reduce style).
• Enhance notes with transcript context.
• Transcript context limit increased from 8K to 30K chars.

​

🚀 New Features

• All 4 quick actions visible in NotchBar expanded view.

​

🔧 Technical

• Update Apple Intelligence dylib with improvements.