Picture an in-person meeting with six people. You clip a small AI recorder to your shirt, the device promises studio-grade capture. Two colleagues sit across from you at the long table, someone mumbles, two briefly talk at once. After the meeting you read the transcript and realise: half of it is inaccurate, and who said what is often unclear.
That is not the weakness of one particular device, but the limit of the "one device, one position" principle. In this article we take an honest look at Pocket, where it is strong, where it hits its limits, and which approach actually solves this problem.
What is Pocket?
Pocket is a wearable AI voice recorder from the US company Open Vision Engineering Inc. in San Francisco. The device is small and light (around 52 grams), clips to your clothing or is worn magnetically, and records conversations in the room. Through a contact microphone it also captures phone calls. In the app this turns into transcripts, summaries, tasks and even mind maps, with over 120 languages supported according to the maker.
On price, the hardware starts around 99 US dollars (regularly 199), with a Pro plan at 19.99 US dollars per month. For what it is meant to do, Pocket is a clever device: personal voice notes, one-on-one talks and phone calls are easy to capture. That is exactly where its strength lies, and where it ends.
The core problem: one device, one position
Every hardware recorder, whether Pocket, Plaud or another device, shares the same physical limit: it records from a single position. As long as only two people speak close together, that works well. In real meetings, reality looks different.
- Distance costs quality. Whoever sits near the device is captured clearly. Whoever sits at the far end of the table comes through fainter and less distinct, so the model has to guess more.
- Overlapping voices. As soon as two people speak at once, the signals blend into one track. Telling apart who said what becomes guesswork.
- Speaker separation without an anchor. A single device attributes speakers purely acoustically. In larger groups with similar voices, that leads to mix-ups in the transcript.
The result is not a bad product, but a limit built into the design. A better microphone or a better model shifts it a little, but cannot remove it.
The other approach: phones become the microphones
Instead of hoping one device covers the whole room, Sally flips the principle. For in-person meetings, several participants' phones can be paired into one shared recording. Each phone becomes a close microphone for its person. You can see how this works in detail on the page about transcription of in-person meetings.
That solves exactly the three problems above:
- Every voice is captured up close. No one sits "too far away" anymore, because each person has their own microphone right in front of them. That is the biggest lever for accuracy, more than any single model.
- Clean speaker separation. Who is speaking is unambiguous through their device, instead of acoustically guessed. In larger groups especially, that is worth a lot.
- No extra hardware. No one has to buy a device. The phones everyone already has are enough.
Honestly said: this does not make any transcript magically "perfect" either. But recording quality in real rooms rises noticeably, especially where single devices struggle most: with multiple speakers and larger distances.
Pocket and Sally head to head
Both capture speech and turn it into text. The difference lies in the recording principle, the data location and what happens after the transcript.
| Feature | Sally | |
|---|---|---|
| In-person recording principle | One worn device | Several phones pairable as close mics |
| Speaker separation in large groups | Purely acoustic | Unambiguous per device |
| Online meeting bot (Meet, Zoom, Teams, Webex) | No | Yes, native |
| Extra hardware needed | Yes, buy a device | No, existing phones |
| Data location | US cloud | Germany, GDPR-compliant |
| Languages | 120+ | 99+ |
| After the transcript | App with summary and tasks | Besides tasks and summary, also chat, search and collaboration |
| Price | from $99 device + $19.99/month Pro | from €8/month, 30 days free |
Data location and GDPR
The second big difference has nothing to do with acoustics. Pocket is run by a US company and stores recordings in its cloud. For European users that usually means the conversation data sits in the US. After Schrems II, transferring personal data to the US remains legally delicate, especially for regulated industries or companies with strict data residency requirements.
Sally is operated by Aliru GmbH in Germany, and processing happens exclusively in Germany. That removes the entire third-country issue. On top of that comes the legal point of any recording: in Germany, Section 201 of the Criminal Code makes secretly recording the non-public spoken word a criminal offence. Recording is only allowed with everyone's consent. A discreetly worn device raises the risk that someone does not even notice the recording; visible, announced documentation creates transparency here.
When Pocket is the right choice, and when it is not
Fair is fair: for individuals who mainly want to capture their own voice notes, one-on-one talks and phone calls, Pocket is a practical device. Anyone dictating quick thoughts on the go gets a lot for little effort.
But as soon as it is about teams working together in changing groups, in person and online, the priorities shift. Then what counts is clean speaker separation with several people, coverage of online meetings, GDPR-compliant hosting and the question of what happens after the transcript. That is exactly where the software approach plays to its strengths, without anyone having to buy hardware per head.
If you know both worlds: much like the topic of Plaud and its alternatives, it pays to look honestly at how often you really sit in small two-person talks in person, and how often in larger groups or online. For the data protection angle, our analysis of hardware recorders and GDPR goes deeper.
Conclusion
Pocket is a clever device for personal recordings and one-on-one talks. Its limit is not a software bug but the principle of the single microphone: in real meetings with several people, accuracy and speaker separation suffer. Pairing several phones into close microphones sidesteps exactly this limit, with no extra hardware and, in Sally's case, GDPR-compliant from Germany.
If your team regularly works together in person and online, try the software approach on one real meeting. You can test Sally free for 30 days and let it join your next appointment.
