Why cloud transcription is awkward in education

Ethics approval, student privacy and shrinking budgets, in that order.

Ethics and participant privacy

A research ethics application that says interview audio is uploaded to a commercial transcription service invites questions about storage, location and secondary use. “Processed on the researcher’s own device” is a much shorter paragraph.

Students didn’t sign up for a bot

Recording a seminar with a notetaker that appears in the participant list makes students the subject of a tool they had no say in. Local capture keeps the decision — and the disclosure — with you.

Per-seat pricing, per-term budgets

Subscription transcription bills by the minute or the seat, which is a poor fit for teaching that clusters into a few intense weeks. One purchase, unlimited local recording.

Where it fits in academic work

Lectures and seminars

Record the room through the microphone, or a hybrid session by capturing both the microphone and the system audio of the video platform. Export SRT or VTT captions to accompany a recording, or Markdown notes for the module page.

Research interviews

Face-to-face or online, both sides are captured — your microphone on one track, the participant’s audio on the other. Because the originals are kept, you can retranscribe with a bigger model when accuracy matters more than speed.

Supervision meetings

Doctoral and dissertation supervisions produce commitments that are easy to forget. Speaker labels, added afterwards by a local diarization model, make it clear who agreed to what before the next meeting.

A term’s worth, searchable

Everything lands in a local searchable library backed by WAV files and a SQLite database. Finding the seminar where a concept was introduced does not require a cloud index.

Fits your own workflow

Markdown, plain text, a JSON archive, SRT/VTT captions and the WAV audio — enough to move a transcript into NVivo-style coding, a reference manager, a VLE or just a folder of files.

Your choice of AI, or none

No cloud AI is involved at any point. A Meeting action copies a summary, action-items or custom prompt together with the transcript to your clipboard, so you can paste it into whichever assistant your institution permits — or, with the optional on-device AI notes switched on, have the note written locally. Or analyse it yourself.

What stays local

Recordings, transcripts and library metadata are stored on your laptop as WAV files plus a local SQLite database. Transcription uses a Whisper-family model that runs on your own hardware, downloaded once and then reused offline. The only network traffic is that model download and licence activation or checkout — no accounts, no telemetry, no participant audio.

Student and participant privacy

Local processing removes one privacy risk — the third-party processor — and leaves the others with you. A recording of a seminar is still personal data about the students in it. Interview audio is still governed by whatever your ethics committee approved.

Because processing happens on-device, JesRecap is designed to support your institution’s compliance process under GDPR-style rules, rather than to certify it. Retention, data minimisation, laptop encryption, who else can unlock the machine, and whether the recording can be shared are decisions your department owns.

Record with consent

Tell students, interviewees and supervisees that you are recording, explain what happens to the file, and follow your institution’s policy and your ethics approval. If a participant declines, don’t record — a quiet local recorder is still a recorder.

Questions

Does interview audio go to a transcription service?

No. Transcription runs locally with a model on your own machine. The only requests JesRecap makes are the one-time model download and licence activation or checkout.

Can I produce captions for a recorded lecture?

Yes. SRT and VTT exports are built in, alongside Markdown, plain text, JSON and the WAV audio.

How accurate are the speaker labels for a seminar with many voices?

Speaker labels are added after recording by a local diarization model and distinguish you from other speakers. They are a reading aid; check them before relying on who said what.

Is there a discount for students or staff?

Pricing is one-time, by device count: €34.95 for 1 device up to €99.95 for 8, with no subscription. For departmental or volume licensing, contact us via the enterprise page.

Will it work in a lecture theatre with no reliable Wi-Fi?

Yes, once the speech model is downloaded. Recording and transcription need no network after that.

Keep teaching and research audio on your own laptop

macOS 14.2+ and Windows 10/11. from €34.95 once, no subscription. The 48-hour free trial is coming soon.