Version 2.1.7 makes recordings easier to review, verify, and recover. - Recent Notes is now the primary Home surface, with clearer Note, Transcript, and Speakers views for recording details. - Personal recordings offer Clean and Raw transcript views. Clean provides a readable transcript while Raw preserves the source transcript for review. - Unresolved speakers now use stable ordinal labels, with recording-level and single-turn speaker corrections in Speaker Review. - Long recordings can resume processing from durable checkpoints after interruption. - Processing recovery is clearer with Open saved recording and Retry processing. - Ask Panora source and citation controls now distinguish verified citations from uncited transcript moments. - Playback controls remain reachable at larger accessibility text sizes.

Version 2.1.8 gives you more control over each new recording and makes long sessions more resilient. • Processing that follows your preferences: Choose Auto, Personal Memo, Dictation, or Meeting; control overviews, action items, decisions, and verbatim cleanup; and add custom instructions. • More reliable long recordings: Transcript playback and search remain responsive across long timelines, live transcription retries silent starts, and chunk transitions show when each part is saved. • Safer recovery: If audio preparation or hosted syncing is interrupted, your local recording stays available. Open the saved recording and retry Panora Processing when ready. • Smoother account access: Continue with Apple directly from On-Device Mode, with clearer offline and session-expired guidance. • Better accessibility: Larger tap targets and clearer VoiceOver support across recording, processing, search, calendar, Privacy Mode, Speaker Review, and Connected Apps.

Version 2.1.5 gives you safer ways to connect Panora with your AI tools and more control over local recording storage. • Connected AI Apps: Create scope-limited, expiring access tokens for Claude Code, Cursor, Codex, and other MCP clients from Settings. Revoke access whenever you choose. • Local audio control: Delete local audio to free up space while keeping the transcript, summary, notes, and search index. • Easier recording review: Skip backward or forward 15 seconds and correct recording title spelling from Manage Recording. • Better name corrections: Save entity and speaker-name corrections more reliably and keep them consistent across recordings. • Safer recovery: Improved handling for iCloud, oversized, empty, and unreadable audio imports, more reliable upload retries, and warnings before signing out with unsynced recordings. • Ask Panora: Ask about past meetings now routes correctly into chat.

What's New — Version 2.1.4 - Personal AI Endpoint model selection: Connect your own AI models — on-device or on your local network — and pick your preferred model right from Settings. - Private, local intelligence: Meeting summaries, action items, and takeaways can be generated entirely on-device, with clearer privacy and hardware guidance in Privacy Mode. - Stronger data protection: Local recordings and their metadata stay on your device — encrypted at rest and excluded from cloud backups. - More reliable uploads: Interrupted audio preparation now recovers and finishes on its own instead of stalling. - Polish and fixes: More consistent design, smoother speaker playback, steadier Home briefs, and clearer error messages throughout.

Version 2.1.3 improves recording reliability and recovery. • Meeting reminders now work better when Panora runs on Mac, with a full-height, scrollable explanation and a pinned primary action. • Stalled or interrupted recordings retain the saved audio and offer recovery. Starting a new recording remains blocked until the capture is recovered or safely discarded. • Imported audio validation now catches video-only MP4s and damaged or incomplete files with clearer guidance. Convertible audio can be normalized to a playable format. • Compact and accessibility layouts are more resilient: Record scrolls instead of clipping content, while Ask context chips and long action buttons remain reachable at larger text sizes.

Version 2.1.2 improves control over AI processing and reliability across recordings, calendars, and Ask Panora. • Panora Processing is now opt-in with a clear hosted-AI privacy disclosure. You can choose On-Device Only. • Local recordings now complete and save their summary, title, and classification after transcription, with retry support and backfill for existing recordings. • Calendar context is more accurate: invalid or ambiguous events no longer add incorrect attendees or recording links. • Personal AI Endpoint routing is more predictable, with no silent fallback to another provider. • Ask Panora and Knowledge Base citations are tied to source evidence, and deleted recording content is excluded from retrieval. • Improved the meeting-reminder acknowledgement screen on smaller displays.

This update adds meeting reminders and improves Ask Panora reliability. • Meeting reminders alert you before eligible calendar meetings without showing meeting names. Tapping opens a prefilled Record screen but never starts recording. • Ask Panora stays responsive after imports and as your source library grows. • Long Ask Panora responses can now be stopped cleanly, with no late answer appearing after cancellation. • Temporary Apple sign-in check failures now show clearer offline, timeout, and retry options. • Hosted Ask Panora now uses DeepSeek V4 Flash for answers.

This update improves processing reliability and speed. • Fresh recordings remain in Processing while Panora sync starts instead of incorrectly showing “Recording unavailable.” • Recoverable on-device audio stays available with a clear retry path if processing cannot begin. • Long, multi-part recordings process faster because transcription and speaker-detection models are reused across audio chunks.

This update improves recording chronology, speaker detection, and processing reliability. • Recordings retain their original capture time after syncing, keeping the Library and recording search in the correct chronological order. • Speaker detection is more consistent across long recordings and less likely to create extra or inconsistent speaker labels. • Fresh recordings stay in Processing while hosted sync starts instead of incorrectly showing “Recording unavailable.” • If processing cannot start, on-device audio remains available with a clear retry path.

This update combines the improvements from builds 46 and 47. • Clearer onboarding and reorganized Settings make permissions, Voice Profile, product analytics, Privacy Mode, and advanced controls easier to understand. • Privacy Mode now blocks new hosted work immediately, pauses optional analytics, and marks affected recordings as “Paused by Privacy.” • Processing percentages appear only when real progress is available, with clearer Waiting, Preparing, and Working states. • Completed recordings recover more reliably after backgrounding, relaunching, or opening a notification, and show Ready only after final results settle. • Recording startup is more reliable on Apple silicon Macs, and long recordings retain their complete speaker timeline during playback. • Ask now supports up to 10 turns per account per UTC day and clearly explains when the allowance resets.

This update makes on-device speaker separation setup more reliable and secure. • On-device speaker separation models now download into persistent storage and remain available across app launches. • Every model package is checked against a signed manifest for its source, expected size, and integrity before use. • Verification details remain stored on-device so installed models can be validated again after relaunch. • Incomplete, altered, or incorrectly packaged downloads are rejected instead of being loaded.

This update improves recording setup, privacy clarity, and processing reliability. • Microphone selection is now always available before recording, including when iOS initially reports only one or no active inputs. • Panora verifies the selected microphone before capture and clearly reports when that input is no longer available. • Failed recording starts now offer “Choose microphone” or “Try again” instead of leaving you at an opaque error. • Analytics, data-access, export, and hosted-processing Settings are easier to scan, with full explanations available through “Read more.” • Hosted-processing consent and Settings now clearly identify Fireworks AI and DeepSeek V4 Flash. • Additional safeguards improve hosted-processing retries, workspace writes, and recovery from service interruptions.

This update makes recording, reviewing, and recovering your conversations more reliable. • Long recordings are now saved in durable parts, reducing the risk of losing audio if the app closes unexpectedly. • Active recordings remain accessible while you move around the app, with a persistent Return to Recording control. • Review uncertain transcript turns and quickly confirm or correct each speaker. • Corrected speaker labels stay intact after syncing, reprocessing, or reopening a recording. • Your existing local recordings remain available while signed out or offline. • Imported audio retains a playable local copy when its original format is supported. • Microphone switching and recovery are more reliable when an input becomes unavailable. • Offline reconnection now provides clearer Retry or Sign In guidance. • Privacy choices, Voice Profile opt-out, hosted AI consent, and Delete All Data handling are more robust. Also includes stability, navigation, and data-integrity improvements throughout the app.