Changelog

October 30, 2026 · 1.19.0

• FIXEDSpeaker recognition never learned voices on work networks — the voice-print model was fetched from Hugging Face, which many corporate networks block, so named speakers were forgotten every meeting. It now downloads from our own mirror like the other models. If Settings → Models shows Speaker recognition as not downloaded, click Redownload.
• FIXEDSettings → Models hid three packs — the live and offline speaker models and Speaker recognition were missing from the list, and the green “models are valid” banner never checked them.
• NEWZoom, Slack, Discord, FaceTime, GoTo and RingCentral meetings are detected — meeting detection knew only Teams and Webex, so a Zoom call never offered to record.
• IMPROVEDA dead headset, a full disk or a failed start can no longer pass as “Recording” — the app watches the audio device every two seconds, reopens it in place when it stops (Bluetooth drop, sleep, unplug), splits into a new meeting if it can’t, and shows a clear error with Retry or Reveal Folder instead of a silent empty recording.
• IMPROVEDClearer start-up — the menu says what it is waiting for while macOS asks for System Audio Recording, and the Request Permission button now works once you have answered.
• IMPROVEDTranscript cleanup can’t hang forever — a stuck cleanup is stopped after a time limit and redone on the next launch, and quitting mid-cleanup no longer leaves two of them racing.
• IMPROVEDBetter logs for speaker names — the app log now records why each voice did or didn’t match a saved person, and the naming popover says when the recognition model is missing rather than blaming the audio.

October 30, 2026 · 1.18.0

• FIXEDIn-person meetings recorded nothing — with no other app playing sound (no Zoom, no browser audio), a recording could capture no audio at all, or only start once something played. Recording now starts right away whether or not anything is playing.
• FIXEDFreeze during long meetings — on macOS 26 the live transcript’s auto-scroll could get stuck in a loop, freezing the app and growing its memory. Auto-scroll now follows only your own scrolling and updates at most a few times a second.
• IMPROVEDRecordings survive a crash or Quit — audio is saved every 2 seconds, so a crash or force-quit keeps everything up to that point. Quitting (or installing an update) during a recording finishes the files first, and an interrupted meeting is transcribed the next time the app opens.
• IMPROVEDStarting a recording no longer freezes the menu bar — the few seconds macOS takes to set up audio capture now happen in the background, with “Starting…” shown meanwhile.
• IMPROVEDBack-to-back meetings — starting a new recording no longer cancels the previous meeting’s transcript cleanup; it finishes in the background.

October 28, 2026 · 1.17.1

• IMPROVEDNew app icon — a graphite headset in the macOS Liquid Glass style.
• FIXEDMeeting detection after sleep — the first call after your Mac woke up could start with no “Record this meeting?” prompt. Simple Meeting Recorder now also checks the microphone every couple of seconds instead of relying only on macOS to report the change.

October 23, 2026 · 1.17.0

• NEWSmarter tag suggestions — Suggest Tags now ranks the tags you already use by how much a meeting resembles the ones carrying them: same speakers, same Clientry client or project, similar topics, same weekly slot. Each suggestion says why, and nothing is saved until you hit Save.
• NEWOffer to tag look-alikes — after you add a tag, Simple Meeting Recorder offers to apply it to similar meetings that are missing it.
• NEWReview Tag Suggestions & Manage Tags — new library-wide sheets to work through suggestions in bulk and to tidy, describe or merge your tags. Dismissed suggestions stay dismissed.
• NEWIndex Voices — one click computes voice prints for older meetings so speaker-based suggestions work on your whole library.
• IMPROVEDAI tags show their evidence — the AI only picks from your existing tags, and each pick quotes the line of the transcript it relied on; invented quotes are downgraded.
• IMPROVEDAgent API sees tag descriptions — list_tags now returns each tag’s description when set.
• IMPROVEDNew website — this site now walks through the app with real screenshots, light and dark, plus a 41-second tour and three short demo videos.

October 14, 2026 · 1.16.0

• NEWAgent API — AI agents like Claude Code can now work with your meetings. They can list your tags, find meetings by tag, date, speaker or text, search transcripts for the exact moment something was said, and read transcripts directly. Connect over MCP (works even when the app is closed) or turn on the local REST server in Settings → Agent API. Everything stays on your Mac.
• IMPROVEDAction items show their source — each extracted item now quotes the line of the meeting it came from, and items whose quote isn’t actually in the transcript are flagged “not in transcript” so you can check them before sending.
• IMPROVEDNo repeat action items — re-running extraction skips items you’ve already sent to Clientry.
• IMPROVEDNewer default AI models — new setups default to the latest models for VibeProxy and Claude.

October 8, 2026 · 1.15.1

• NEW“Me” filter for action items — one click shows only the items assigned to you. While it’s on, Set All Destinations and Send only touch the items you can see.
• IMPROVEDEvery action item needs a destination — the “Inbox” option is gone. Items without a Clientry project are highlighted and never sent, so nothing lands in Clientry unfiled.

October 8, 2026 · 1.15.0

• NEWMeeting links in Clientry — action items you send to Clientry now link back to the meeting they came from. Click the task’s Source link and Simple Meeting Recorder opens that meeting, scrolled to the moment the item was agreed, with playback ready to start right there.
• IMPROVEDAction items know when they were said — extracted action items now remember the point in the meeting where they came up.
• FIXEDPasted API keys — an AI provider key pasted with stray quotes, a trailing comma or extra spaces (for example, copied out of a JSON file) now works.

October 6, 2026 · 1.14.2

• NEWTag editor — Edit Tags is now a proper panel: add tags as you type with suggestions from the tags you already use (most-used first), tap a common tag to add it, remove tags with the ×, and use Suggest with AI without leaving the panel.
• IMPROVEDBolder tags — tags in the meeting list are now solid, high-contrast pills that stay readable on the selected row and over remote screens.
• FIXEDSlow startup — the app no longer waits on the audio system before showing the menu-bar icon and Meetings window; meeting detection now starts quietly in the background.

October 1, 2026 · 1.14.1

• NEWScreen-share timer setting — Settings → Recording → Menu Bar has a switch for hiding the recording timer while you share your screen. Turn it off if the timer next to the red icon keeps disappearing and reappearing when you aren’t sharing.
• FIXEDScreen-share detection — invisible full-screen windows from browsers and meeting apps no longer count as a screen share.

October 1, 2026

• NEWTags — tag meetings from the tag button or by right-clicking a meeting. Tags show on each meeting and in the list; click a tag to see every meeting with it, click its × to remove it, or pick one from the tag menu above the list. Search finds tags too.
• NEWAI tag suggestions — choose Suggest Tags and the AI proposes a few tags from the transcript, reusing tags you already have. Suggestions are added to your tags, never replacing them.
• IMPROVEDMeeting prompt — the record card slides in gently, its buttons always fit on one line, and the wording is clearer about which call started or ended.

September 24, 2026

• NEWMeeting prompt that stays out of your way — when Teams or Webex starts a call, a small card offers to record in the corner instead of a pop-up that steals focus. It never shows up in a screen share, and closes itself if you ignore it.
• NEWAI meeting titles — right-click a meeting (or open Rename) and choose Generate Title. Turn on Settings → AI → Auto-title new meetings to name each meeting when its transcript is ready. Off by default; titles you set yourself are never replaced automatically.
• NEWPlay from any point — click the ▶ timestamp on a turn to play from there. The turn being played is highlighted and followed, and Space plays or pauses.
• IMPROVEDEasier-to-read transcripts — one bubble per speaker turn instead of many fragments, with timestamps that match the player.
• IMPROVEDMeetings list — grouped into Today, Yesterday, This Week, and by month, with each meeting’s length and speaker count. Each meeting now shows its date, length, and speakers at the top.
• IMPROVEDRecording timer while sharing — the menu-bar timer hides while you’re sharing your screen.
• IMPROVEDVoiceOver — toolbar and chat buttons are announced by what they do.
• FIXEDModels status — Settings → Models no longer says the models are valid before it has checked them.

August 27, 2026

• NEWApp log — Settings → General → Diagnostics lets you reveal or copy a support log. Recording, transcription, models, and AI write here; meeting text and API keys are not included.
• NEWClaude — Settings → AI can use Anthropic’s Messages API (official Claude or any host that speaks it), with its own URL, key, and model.
• NEWTranscription models — Settings → Models lists the on-device packs, checks them with SHA-256, and can redownload if a file is missing or corrupt.
• IMPROVEDModel validation — Validate and redownload now say which files passed or failed, with progress while a pack is downloading.
• FIXEDModel folders — checksum checks find packs even when FluidAudio nested the cache one folder deeper.

August 13, 2026

• FIXEDAI model list — switching Settings → AI from OpenAI to VibeProxy (or back) now reloads the models from the new backend instead of leaving the previous host’s list.

August 6, 2026

• NEWOpenAI-compatible AI — Settings → AI now lets you pick VibeProxy or OpenAI. Point chat and action items at api.openai.com or any local OpenAI-style server (omlx, LM Studio, and similar), with an API key when the host requires one.
• IMPROVEDLocal AI hosts — paste a base URL with or without /v1; both work. Longer timeouts for cold model loads, and better handling of thinking-model streams.

August 3, 2026

• IMPROVEDRecording timer — the live elapsed time in the menu bar (and meeting views) uses a clean monospaced font so digits stay aligned as seconds tick.

July 31, 2026

• IMPROVEDInstall without admin — first-run install copies into your user ~/Applications folder so no administrator password is required. Automatic updates still work.
• NEWReprocess transcript — re-run offline transcription and speaker labels on any past meeting from the toolbar, sidebar menu, or empty-state button. Progress shows in the library; a failed pass never blanks your existing transcript.
• NEWStop when the call ends — after you start recording a Teams or Webex meeting, SMR watches for the call to leave and offers to stop so you don’t keep rolling into empty air.
• IMPROVEDSettings uses a System Settings–style sidebar (General, Recording, AI, Clientry) with clearer permission status.
• IMPROVEDLive transcript builds turns from speech-token timing, so overlapping speech and mid-sentence corrections land more reliably while you record.
• IMPROVEDLower CPU and UI overhead during long recordings (less work per spoken token).
• IMPROVEDChangelog is easy to find in the site nav and footer.
• FIXEDReprocess / polish crash — offline transcription runs in a separate worker so a bad audio decode cannot take down the menu-bar app; also fixed a frame-count trap in the audio converter.
• FIXEDEdge cases that left an empty chat bubble, mis-clamped audio channels, or drop-related flicker in the live transcript.

July 21, 2026

• NEWOpen-speaker echo cancellation — when remote audio plays from your speakers, it no longer shows up as a second “Me” voice in the transcript. Auto mode turns this on for built-in speakers and off for headphones or Bluetooth.
• NEWRecording preference in Settings: Echo cancellation Auto / On / Off.
• NEWAbout panel from the menu bar (and the app menu when a window is open).
• IMPROVEDLive and polished transcripts drop residual speaker-bleed lines that slipped past cancellation.

July 13, 2026

• NEWAction items — AI-extracted action items from every meeting, with one-click send to Clientry.
• NEWSettings window (Cmd-,) for configuring integrations.
• IMPROVEDTranscription models now download from our own servers, so first-run setup works on networks that block huggingface.co.

July 7, 2026

• NEWAI chat — ask questions about your meetings, answered entirely by a local LLM. Nothing leaves your Mac.
• IMPROVEDLive transcription now shows up right in the meeting library as an in-progress entry that updates in place when the meeting ends.
• IMPROVEDNew app icon, and the transcript window now shows a dock icon.
• FIXEDFixed an auto-update error affecting development builds.
• NEWNew shorter address: smr.dmalson.com