Your meetings never become someone else's data
Data residency is usually a long conversation about regions, contracts, and subcontractors. When the recording and the transcription happen on the machine in front of you, most of that conversation is already over.
The data stays where it was made
Audio, transcript, speaker labels, and the searchable library are written to the same device that recorded the meeting. Whatever jurisdiction that laptop is in, that is where the data is.
No copy on our side
We do not receive meeting content, so there is no vendor database holding it, no retention setting outside your control, and nothing for us to hand over to anyone.
Deletion is an action, not a request
Removing a meeting is a file operation on your own disk. No support ticket, no waiting for a deletion window, no wondering about backup copies you cannot see.
Why residency is hard with cloud assistants
This is not a criticism of any particular product. It is the architecture: if a service transcribes for you, the recording has to get to the service.
Cloud transcription
The recording is uploaded, stored, and processed by a provider. Residency then depends on which region you were given, where the model inference runs, where support staff sit, and which subcontractors are in the chain this quarter. Each of those is a claim you have to verify from outside, re-verify when terms change, and describe in your own documentation. Deletion becomes a promise. If a bot joins the call to record it, the meeting platform is in the chain too.
JesRecap
Capture runs through the operating system's audio APIs; transcription runs against a model file on the local disk. Nothing about the meeting is sent anywhere, so there is no region to choose, no subcontractor list to track, and no transfer mechanism to paper. The residency answer is the location of the device — a fact you already know. No bot joins the call.
What "on-device" means concretely
Positioning pages get vague at exactly the point where a reviewer needs detail, so here is the detail. JesRecap records two separate tracks — your microphone and the audio your computer plays — and mixes a playback track afterwards. Transcription uses a Whisper-family model downloaded once and then reused from disk. Speaker labels come from a local diarization model. Meetings live in a per-user folder with a SQLite index and WAV files, and stay there until you delete them.
The complete list of network traffic the product involves: a one-time speech model
download from models.jesrecap.com, licence activation and validation,
and a version check. No telemetry, no analytics, no accounts. The
security overview lists it endpoint by endpoint, and
the deployment guide explains how to
pre-seed models so even the download does not happen on the user's machine.
One honest caveat. Sovereignty ends where you choose to move the data. JesRecap has no cloud AI, but its meeting actions can copy a prompt with the transcript so you can paste it into whichever assistant you already use. That paste is an upload, and the residency question comes straight back for that vendor. If you would rather it never arose, turn on the optional on-device AI notes and the summary is written by a model on your own machine — no processor, no jurisdiction, nothing to map. Either way the decision is yours, rather than us routing your transcripts through a service you did not pick.
Where this matters most
The same architecture, seen from four different review processes.
Public sector
Residency is often a hard requirement rather than a preference.
Read more →Healthcare
Patient detail in the audio, and a vendor list you have to justify.
Read more →Legal
Privileged conversations that should not be stored by a third party.
Read more →Financial services
Client calls, plus supervisors who ask where records are held.
Read more →The residency answer is "this laptop"
Record, transcribe, and keep everything on your own machine. from €34.95 once, no subscription. macOS 14.2+ and Windows 10/11.