On-device transcription for clinical conversations
Consultations, telehealth appointments, handovers and team meetings — recorded and transcribed on the clinician’s own machine. Nothing is uploaded, so there is no new processor to approve before you can start.
The problem with cloud notetakers in care settings
Not that they transcribe badly. That the recording travels.
The audio leaves the practice
A cloud assistant sends patient voices to a server you don’t run, in a country you may not have chosen. Every approval conversation starts there and rarely gets easier.
A bot in the appointment
Notetakers that dial into a video consultation appear as an extra participant. Patients notice, and asking a distressed person to accept an unfamiliar attendee is its own harm.
Retention you don’t control
Once audio is on someone else’s infrastructure, deletion is a policy promise rather than an action you can take. Local files you can move, encrypt with your disk encryption, or delete outright.
How JesRecap fits a clinical day
The same app covers the room and the screen, because it records two different things.
In person, through the microphone
Put the laptop on the desk and record the consultation as room audio. No meeting platform is involved at all — JesRecap is simply capturing your microphone input.
Telehealth, through system audio
For video appointments JesRecap records the audio your computer already plays, using Core Audio process taps on macOS or WASAPI loopback on Windows. It works with Zoom, Teams, Meet, Webex or any other app, and nothing joins the call.
Afterwards, on the same machine
A local Whisper-family model produces the transcript, a mixed playback track is written for review, and you export Markdown, plain text, JSON, SRT/VTT captions or the WAV audio into the patient file or your own notes.
Speaker labels, added locally
After a recording finishes, a local diarization model separates “You” from the other speakers, so a consultation transcript reads as a conversation rather than a wall of text. The model is a one-time 54 MB download and then runs offline like the rest. It is a useful reading aid, not an identity check — review labels before anything goes into a record.
Two tracks, kept
Your microphone and the system audio are written as separate WAV files and only mixed during post-processing. That keeps the clinician’s side and the patient’s side distinguishable, and it means you can retranscribe an old appointment later — with a larger model, for example — from the original audio.
What stays local
Audio, transcripts and metadata stay on the device, stored as WAV files plus a local SQLite database in your user library folder. Speech recognition and speaker labelling both run on your own hardware. The only network traffic JesRecap makes is the one-time model downloads and licence activation or checkout. No accounts, no telemetry, and no meeting content in either request.
Summaries are the one place people expect a cloud call, and there isn’t one. A Meeting action copies a prepared prompt with the transcript to your clipboard, and you decide whether to paste it into an AI tool your organisation has already cleared, or to keep the transcript as it is. If you would rather no tool were involved, switch on AI notes on this device and the summary is written by a model on the same machine — a summary with no data processor attached, and no business associate agreement to negotiate for it. The setting is off until you turn it on.
Being careful about compliance
JesRecap is not “HIPAA certified” and we will not describe it that way. No recording app can be, on its own: compliance is a property of your organisation, your policies and how you use your tools. What software can do is remove steps that are hard to justify.
Because processing happens on the device, JesRecap is designed to support your compliance process under GDPR- and HIPAA-style requirements — the situations where cloud transcription is difficult to approve because it introduces a third-party processor, a cross-border transfer, or a retention policy outside your control. There is no vendor-side copy of a consultation to inventory, and no data processing agreement needed for a transfer that does not happen.
The parts that remain yours: lawful basis and consent, how long you keep recordings, disk encryption and device security, who can unlock the machine, and whether a recording belongs in the clinical record at all. Ask your privacy officer to review the specifics — we would rather hand them an accurate description of a local pipeline than a badge.
Record with consent
Tell patients and colleagues that you are recording, before you start, and follow the consent rules of your jurisdiction and professional body. A quiet local recorder is still a recorder. If someone declines, don’t record.
Questions
Does any patient audio reach your servers?
No. Recording, storage and transcription all happen on the device. The only requests JesRecap makes are the one-time speech model download and licence activation or checkout, and neither carries meeting content.
Can I use it for in-person consultations, not just video calls?
Yes. In-person is the simpler case: JesRecap records the microphone track and no meeting platform is involved. Telehealth adds the system-audio track so both sides of the call are captured.
Will patients see a recording bot in the appointment?
There is nothing to see. JesRecap records the audio your computer already plays instead of joining as a participant, so the attendee list is unchanged. You should still say out loud that you are recording.
Does it write the clinical note for me?
No, and it does not pretend to. JesRecap produces a transcript and can copy a summary or action-items prompt with the transcript to your clipboard for whichever AI you already use. Drafting and sign-off stay with you.
Where are recordings stored, and can I delete them?
In your application support folder as WAV files with metadata in a local SQLite database. You can export, move or delete them like any other file on the machine, and your existing disk encryption applies.
Related use cases
Same architecture, different room. Or see all industries.
Transcription that never leaves the consulting room
macOS 14.2+ and Windows 10/11. from €34.95 once, no subscription. The 48-hour free trial is coming soon.