How AI memory pendants work

A short definition

An AI memory pendant is a small wearable that listens to speech, converts it to text, and uses software to turn that text into something you can search or act on: a summary, a list of tasks, a searchable archive. The word pendant covers a lot of shapes in practice, from chest-worn jewelry to clips and wristbands. What ties them together is the pipeline behind the hardware. Legend Memory's Archive explains its version of the definition in What is a memory pendant?, which is worth reading alongside this page.

The pipeline in four steps

1. Capture

One or more microphones pick up speech. More microphones extend the pickup range and help separate speakers. Plaud's Note Pro uses four MEMS microphones with a stated 16.4 foot range. The NotePin S uses two, with a stated 9.8 foot range. Legend lists three microphones with speaker separation. Pocket lists two studio microphones plus a contact microphone that picks up phone calls when the device sits on the back of a phone.

Capture is also the point where devices split into two models. Press-to-record devices start when you press a button. Always-on devices start when you switch them on and keep going until you stop them or the battery ends.

2. Transcription

Speech becomes text. Where this happens differs. Plaud and Pocket sync audio to their apps and cloud accounts, where transcription runs. Anker's soundcore Work stores audio encrypted on the device by default and uploads audio to the cloud only for transcription. Legend Memory says Legend transcribes on the device itself and syncs only text to the phone. Where transcription runs decides what leaves the device, which is why the privacy page looks at it closely.

3. Summarization and structure

Large language models turn the transcript into summaries, action items and topics. Plaud says its intelligence layer runs on leading models from several providers. Pocket describes itself as model-agnostic. Summaries are where paid plans usually start: Pocket's free plan limits summary storage to 30 days, and Plaud's paid plans raise the minute allowance and add features.

4. Retrieval

Once captured, the material is only useful if you can find it again. Most products offer a search or a chat interface over your history. Plaud calls this Ask Plaud, Pocket calls it Ask Pocket, and Omi says you can ask it about anything it heard. Limitless exposed the same idea through a developer API for its Pendant, documented on its developer page.

Battery, charging and sync

Battery figures are measured differently by each maker. Plaud lists up to 20 hours of continuous recording for the NotePin S and 40 days of standby. Omi lists 10 to 14 hours. Legend Memory lists up to 2 days of continuous use and 5 days of standby, with USB-C charging in about 2 hours. Anker lists 8 hours for soundcore Work, extended to 32 hours with its charging case, and notes these are internal lab figures. SwitchBot lists up to 20 hours of recording. Bee lists up to 7 days.

Sync matters as much as battery. Pocket and Omi record offline and sync later. soundcore Work auto-launches its app to upload files when a recording stops. Legend Memory says its device holds text locally and sends it to the phone when back in Bluetooth range. If you spend long stretches away from your phone, check how each product behaves.

Why most of these have subscriptions

The hardware is a microphone and a battery. The expensive part is running speech models and language models on your recordings, and that cost recurs every time you capture something. That is why nearly every product here has a paid plan or a free-minute cap. TechCrunch reported in June 2026 that Plaud's chief executive said nearly half of its device owners upgraded from the basic plan, and that this drives much of its revenue. Products differ in whether they charge by the minute, by the month or as one membership. Legend Memory chose one flat annual membership with no minute meters, and says so; Omi and Pocket list free plans. Which model is best for you depends on how much you talk and how much you value simplicity over price.

Why the form factor matters

Microphones hear best when they are close to the speakers and unobstructed. A pendant sits at chest height, roughly between the wearer and the person they are talking to, which is why several makers use it. A clip on a collar is close to the mouth. A wristband points wherever the arm is. A card attached to a phone captures phone calls well and in-person speech less well. None of this is settled science, and we have not seen a neutral test that compares them, so read form-factor claims as design intent rather than measured results.

Form also affects how other people react. A visible light or a clearly worn pendant tells people something is active. A small clip is easy to miss. That matters for consent, which we cover in the privacy and consent page.

What these devices are not

• Not medical devices. None of the products here claims to diagnose or treat a condition. The National Institute of Mental Health describes ADHD as a neurodevelopmental disorder with standard treatments including medication and psychosocial interventions. A capture device is a memory aid at most.
• Not perfect transcribers. Names, numbers and jargon are the usual failures. Devices that keep audio let you replay a moment to check. Devices that keep only text cannot.
• Not a substitute for consent. Wearing a device with a light on is not the same as everyone in the room agreeing to be recorded or transcribed.

What the research says about why capture is hard

Prospective memory, remembering to do something later, is a documented difficulty for some adults with ADHD. A 2013 PLOS ONE study compared 45 adults with ADHD to 45 matched controls and found a large deficit in planning multi-step intentions but little difference in plan recall or in starting an action at a set cue. The authors stress that the problem was planning, not forgetting the plan. That result is a reason for modesty. It does not show that any wearable improves outcomes, and no product on this page has been tested in that study. See Fuermaier et al., PLOS ONE (2013): Complex prospective memory in adults with ADHD.