AI meeting-notes devices are dedicated hardware: built-in microphone arrays capture meeting audio, and AI transcribes in real time, distills summaries and generates action-item lists, with some offering live translation. Versus phone apps, they sell "works out of the box", offline capability and privacy. Representative products include the credit-card-sized Plaud Note, the collar-clip Plaud NotePin, the pendant-style Limitless Pendant 2 ($199, 8 hours of continuous recording) and the Mem Ring, with entry prices mostly in the $100–300 range and claimed transcription accuracy above 95%. See our two category roundups: the March overview and the earlier February survey. The global AI meeting-notes market was about $4.7B in 2025, projected to exceed $12B by 2030.
Plaud is the breakout of this category: by June 2026 it had shipped over 2 million devices, and its software ARR crossed $100M — going from zero to $100M in under 18 months, a journey that took benchmark player Otter.ai nearly six years (full report).
The model is "device as the entry point": every ~$150 device is a subscription-conversion touchpoint, with ARR per shipped unit at about $50 and room to grow. The risks are equally clear: stronger phone and cloud AI could erode the need for dedicated hardware; free ecosystem features like Microsoft Copilot in Teams and Zoom's AI Companion squeeze independents; and the sensitivity of meeting data makes privacy existential.
The Halliday G2 smart glasses, released in July 2026, represent a counterintuitive route: drop the camera entirely and "listen, don't look" — a microphone array and edge AI chip identify multiple speakers and generate structured meeting minutes in real time, a dedicated NPU sustains 8 hours of continuous listening, and voice data is processed on-device by default, with raw audio never leaving the glasses (launch report).
The "downgrade" is precisely the pitch: no camera means no filming accusations by design — HR need not worry about photographed documents, and the device stays usable in confidential meetings and in jurisdictions that ban non-consensual recording. The cost is giving up visual assistance like navigation and overlays — AI wearables are splitting from "do everything" toward "nail one scenario".
Mainstream AI meeting tools are almost all subscription-based and cloud-processing. Two subscription-free options we have covered chart another path: Meetily is a free, open-source transcription-and-summary tool that captures audio locally and generates text and to-dos locally, with auditable code — attractive for lawyers, doctors and financial advisers; Talat takes the local-first route, running transcription and summarization fully offline with notes stored locally as Markdown/JSON, claiming 95%+ accuracy.
A practical caveat: open source converts money costs into learning and maintenance costs — rougher interfaces, hardware requirements, and possibly your own API keys for stronger models. Best for hands-on users and IT teams.
The pure-software camp is firing just as hard: Zoom's AI Companion is free, Microsoft Copilot is deeply embedded in Teams, and in August 2026 Calendly entered with its assistant Callie — bundling scheduling with notes for its ~20M monthly users: pre-meeting prep, live capture, and post-meeting action items synced to the calendar.
When choosing, run four checks: 1) where the data goes — on-device or cloud; for cloud, read the encryption and retention policies; 2) ecosystem fit — does it integrate with your Zoom/Teams/calendar/CRM; 3) true cost — hardware price plus subscription, compared against free built-in platform features; 4) compliance — most jurisdictions require consent for recording, especially in cross-border meetings. Dedicated devices earn their keep in cross-platform use, offline settings and physically isolated privacy; if your meetings are all online, try the free options before buying hardware.
There is no single answer — it depends on your scenario: for cross-platform, offline meetings with out-of-the-box ease, the Plaud line (the category leader with 2M+ units shipped); for camera aversion and compliance, the camera-free Halliday G2; for tight budgets or data that must never leave the device, local open-source options like Meetily and Talat; and if all your meetings are online, built-in AI in Zoom/Teams or Calendly's Callie may be enough.
Dedicated hardware generally charges twice — device plus subscription; that model is exactly how Plaud pushed ARR past $100M. Two ways to avoid it: free open-source Meetily or local-first Talat (no subscription), or free platform features like Zoom's AI Companion. The open-source trade-off is setup effort and self-maintenance.
It depends on jurisdiction and consent: most places require at least one-party or all-party consent, and non-consensual recording can be illegal. Devices with on-device processing (Halliday G2, Talat) reduce data-leak risk but do not waive the consent requirement — clear legal and HR policy before deploying in company meeting rooms.
That is the category's real existential question. Dedicated devices defend three moats: cross-platform and offline coverage (no dependence on one meeting app), long-endurance dedicated capture (mic arrays plus 8-hour-class listening), and physically isolated privacy (audio never leaves the device). If your meetings all happen in Zoom/Teams, the built-in AI is probably enough.
