Does the Legend pendant record audio? No. Legend converts speech to text instantly, as you speak. There is no recording step, which means no audio file is ever written, not even briefly. That is a different claim from the one most privacy pages make. Plenty of devices record and then delete. Bee, for one, publishes that conversations are processed in real time and immediately deleted, never saved or stored. Legend never creates the file in the first place, so what exists after a conversation is a transcript and nothing else.

Device Creates an audio file Price
Legend No, transcribes live $499 USD including the first year of membership, $1,497 over three years
Plaud NotePin S Yes, stored $179, Pro $99.99/yr
Bee Bee states it processes in real time and deletes immediately $49.99, published by Bee
Friend 2.0 Yes, 30-day memory window $249, $10/mo optional

The rest of this piece is what sits behind that first column, and why the distinction is worth more than it sounds.

What Is the Difference Between Recording and Transcribing?

A recorder writes an audio file. That file can be played, copied, emailed, subpoenaed, leaked in a breach, or handed to whoever buys the company. It holds voices, tone, background conversation, and everything else the microphone reached.

A transcriber turns sound into text as it arrives. The words are what get written down. The audio is never a file at any point, so there is nothing to store, protect, or later delete.

That distinction matters more than it sounds. "We delete your audio" invites obvious follow-up questions: deleted when, from how many backups, and what happened during the window before deletion. "No audio file is created" does not have a window. It is the difference between a promise about housekeeping and a fact about architecture. Applied across the category, that test sorts devices into three groups, and we set out which devices actually qualify.

Bee is the clearest published example of the deletion version of that claim. Its Bee Pioneer page states that conversations are processed in real time, immediately deleted after processing, and never saved or stored. That is a meaningfully stronger position than keeping an archive, and it is also not the same position as Legend's. Bee's is a statement about what happens to audio after it has been handled. Legend's is that no audio file is written at any point, so there is no deletion step to trust and no window to ask about. Worth noting too that Bee's strong wording appears on that product page rather than in its Privacy Notice, which speaks more generally about collecting audio or call recordings, so this is Bee's published claim rather than something a buyer can verify from outside.

What Actually Happens to Your Voice?

Three stages, not four, and the absence of a fourth is the point.

Capture. Three microphones pick up speech from a few metres away, with speaker separation so the transcript reads as a conversation rather than one undifferentiated block of words. Legend is worn at chest height, which sits roughly between your mouth and the mouth of whoever is in front of you.

Transcribe. Speech becomes text in real time across more than 150 languages. This happens as you talk, live, not as a batch job run against a saved file afterward.

Deliver. The Legend App on iOS and Android shows the transcription as it happens, then assembles daily summaries and auto-generated to-dos from what was said.

There is no stage where a recording sits somewhere waiting to be processed, and no stage where one gets cleaned up.

Can Anyone Play Back What I Said?

No, because no audio file exists to play. That applies to you as well, and it is worth being straight about it.

Some people will read that as a limitation, and for certain uses it genuinely is one. A journalist checking a quote word for word, or anyone who needs an audio record for a formal process, needs an actual recorder and should buy one. Legend is not built for that job and will not do it.

What you get instead is text you can search and ask questions of, which is the more useful format for what most people actually want, which is to remember what was agreed rather than to relive how it sounded.

What About the People Around Me?

Two things are working here, and both are deliberate.

The first is that nothing anyone else said becomes an audio file either. Their voices are not sitting in an archive anywhere, which is the objection most people actually have when they notice a microphone in a room.

The second is the Halo light, which shows when Legend is actively transcribing. Google Glass failed in 2013 largely because bystanders could not tell what it was doing, so they assumed the worst. A visible indicator is the cheapest honest answer to that, and it turns the device from something that might be listening into something that is plainly either on or off.

The social side of this is a skill of its own, and we went through it properly in Is It Creepy to Wear an AI Pendant?

How Does This Compare to Other AI Pendants?

Device Capture Creates an audio file Visible indicator
Legend Always-on No, transcribes live Halo light
Plaud NotePin S Press to record Yes, stored You started it
Bee Always-on Bee states it processes in real time and deletes immediately Limited
Friend 2.0 Always-on Yes, 30-day memory window Limited
Reflecti Wake word or tap Yes, encrypted segments stored Not stated

Worth being fair about this. Storing audio is not a failure of care, it is a design choice with real benefits: you can verify a transcript against the original, reprocess it later with a better model, and recover from transcription errors. Plaud stores audio because it is a meeting recorder, and that is what meeting recorders should do.

The trade is that you then own an archive of recordings that has to be protected, exported, and eventually disposed of. Bee sits between the two positions rather than at either end, since by its own published account it neither keeps an archive nor avoids handling the audio in the first place.

Can I Delete My Data?

Yes. You can remove any single conversation or wipe your entire history from the Legend App. Data is encrypted at rest and in transit.

The broader point is that the less you hold, the less there is to lose. Companies in this category get acquired and shut down with very little warning, and when that happens whatever archive you built goes wherever the assets go. We wrote about what that has actually looked like in practice in What Happens to Your Data When an AI Wearable Company Shuts Down?

Why Does This Matter Legally?

Because wiretapping statutes were written around intercepting and retaining audio, and a device that converts speech to text live without ever writing an audio file occupies genuinely different territory from one that stores playable recordings. Courts have not settled how the two compare, because the hardware is newer than the case law.

That is not a claim that transcription puts you outside the law. Federal law in the United States follows one-party consent, meaning you can capture a conversation you are part of, and twelve states require everyone present to agree regardless of what format the capture takes. If you live in one of those, the etiquette matters whatever your device does. The full state-by-state picture is in Are AI Wearables Legal?

What the design does change is the aftermath. Nothing exists to be requested, leaked, or handed over except text.

The Takeaway

Legend does not record. It listens, converts speech to text live as you speak, and gives you back notes. No audio file is created at any point in that process.

If you need a playable recording, buy a recorder, and Plaud is the best one in this category, though it is worth weighing against the other serious contender first in Plaud vs Pocket. If what you want is to stop losing what people tell you without accumulating a library of everyone's voices in the process, that is the specific problem Legend was built for. It is a voice-first AI pendant made for people with ADHD, worn at chest height, transcribing in real time across more than 150 languages with three microphones and speaker separation, showing a Halo light whenever it is active, and delivering the day as summaries and auto-generated to-dos in the Legend App. About two days of continuous use from a two-hour USB-C charge, splash resistant, encrypted at rest and in transit, $499 including the first year of membership with a 30-day money-back guarantee. Full specifications and colour options are on the Legend One product page.

FAQ

Does the Legend pendant record audio? No. It transcribes speech to text instantly as you speak, and no audio file is created at any point. What exists afterward is a transcript.

Can I listen back to a conversation on Legend? No. Because no audio file is created, there is nothing to play back. You get a searchable transcript instead, which you can read and ask questions of.

Is that different from deleting the audio afterward? Yes, and it is the stronger position. Deleting means a file existed and then was removed, which raises questions about timing and backups. Legend never creates the file. Bee, for example, publishes that it processes conversations in real time and deletes them immediately, which is a deletion claim rather than a claim that nothing was ever written.

Can Legend staff hear my conversations? There is no audio for anyone to hear. Data is encrypted at rest and in transit, and you can delete any conversation or your entire history from the Legend App at any time.

Do people around me know it is on? Yes. A Halo light shows whenever Legend is actively transcribing, so it is visibly either running or not.

0 comments