It reads the paper, not just the page
Uploaded PDFs are broken into sections, questions, sub-questions, multiple-choice options and mark allocations — so a learner can jump to Question 4.2 instead of scrubbing through audio.
AF + EN Offline Afrikaans voice bundled — no cloud account needed
The exam paper, read aloud — in English and Afrikaans.
Exam Reader is Windows software for South African schools. A teacher uploads the paper, checks anything the system flags, and assigns it. The learner puts on headphones and hears the paper in a locked full-screen reader.
Grade 4–12 · Runs on your school's own server · No learner data leaves the building
Gr 4–12
Papers in English & Afrikaans
On-premise
Your server, your learner data
Under 3 s
Target time from Play to audio
Offline
No internet needed on exam day
What it does
A reader who never gets tired, never paraphrases, and never reads ahead
Learners who qualify for a reader as an exam accommodation depend on a person being available, consistent and neutral for the full paper. Exam Reader does that job the same way every time — and frees the staff member up.
It knows which language it is reading
Every segment is tagged English or Afrikaans automatically, so a bilingual paper is read in the right accent and pronunciation throughout — without the teacher tagging anything by hand.
Nothing is read until a teacher approves it
Images, maps, data tables, matching columns and any low-confidence text are flagged for review. The teacher writes what the learner should hear, and only then does the paper become assignable.
A locked reader, not a web browser
When the exam starts, the client goes full screen and stays there: play, pause, repeat, previous, next, speed and progress. No papers to edit, no learner records, no way to wander off.
Everything is logged
Uploads, edits, approvals, assignments and every playback event during the exam are written to an audit log — the record you need if a result is ever queried.
Built for a room of learners
One school server, the same client app on every teacher and student PC. Learners are imported from Excel or CSV, and the invigilator can watch active sessions from the dashboard.
How it works
From PDF to headphones in four steps
Upload the paper
The teacher signs in and uploads the exam PDF. The server splits it into readable segments and detects the language of each one.
Review what is flagged
Only the uncertain parts need attention — diagrams, tables and matching exercises. The teacher writes the spoken description and signs the paper off.
Assign to candidates
Audio is generated once and cached on the server, then the paper is assigned to the learners who are approved for a reader in that sitting.
The learner presses play
The learner enters their candidate number, confirms the paper, and the app locks into full-screen reader mode for the rest of the sitting.
Voices
The Afrikaans voice is the part nobody else solves
Built-in Windows narrators and generic screen readers do not handle an Afrikaans exam paper well. Exam Reader ships with a neural Afrikaans voice that runs on the school server itself — installed by the setup program, with no subscription and no internet connection required.
- Offline neural Afrikaans and South African English voices, included in the installer.
- Optional Azure Neural TTS if your school already has a subscription and wants the cloud voices.
- Abbreviations, Rand amounts and mark allocations are expanded so they are spoken the way a person would say them.
- Audio is generated once per paper and cached, so thirty learners pressing play together does not queue.
AF
Afrikaans (af-ZA)
Offline neural voice · bundled with the server
EN
English, South Africa (en-ZA)
Local accent, not a US or UK narrator
☁
Azure Neural TTS
Optional upgrade · your own subscription key
Emergency fallback
Built-in Windows voice, if audio was never cached
Data & exam integrity
Learner data stays on the school premises
Exam Reader is not a cloud service. The school installs one server, and candidate records, uploaded papers, teacher-approved text and generated audio all live on that machine — which makes the POPIA conversation with your governing body a short one.
Hosted by you
One Windows server on the school LAN. No tenant, no third-party database, nothing to delete from someone else's cloud.
Works without internet
Papers, voices and OCR all run locally. The LAN needs no internet access unless you choose the optional Azure voice.
Auditable sessions
Every upload, approval, assignment and playback event is timestamped, so an exam session can be reconstructed afterwards.
Backups school IT can run
Weekly automatic backups of the database, papers and audio, plus manual backup and restore from the dashboard.
What you need
Ordinary school hardware
No GPU, no new server room. A four-core machine with 8 GB of RAM and an SSD runs the whole school; the learner PCs only display a page and play cached audio. Headphones per learner are the one thing worth budgeting for.
| School server | Windows 10/11 or Server, x64 · 4 cores · 8 GB RAM · 50 GB SSD |
|---|---|
| Learner PC | Windows 10/11 x64 · 4 GB RAM · headphones · Wi-Fi or LAN |
| Concurrency | Tested against 25–35 learners reading at the same time |
| Internet | Not required, unless you enable the optional Azure voice |
Questions
The things schools ask first
It replaces the reading, not the supervision. A teacher still prepares and approves the paper beforehand, and an invigilator still supervises the room. What changes is that a staff member no longer has to sit beside each learner reading the paper aloud for the full sitting.
Most of a paper is extracted and read correctly without any intervention. The teacher's time goes into the flagged items — diagrams, maps, data tables and matching columns — where a human has to decide what the learner should hear. Our own benchmark for a sustainable workflow is under 30 minutes for a Grade 4–9 paper and under an hour for Grade 10–12.
The server runs OCR over the page images to pull out any text inside them, and flags the visual for teacher review. The teacher writes the description the learner will hear. Nothing visual is described to a learner without a person having approved the wording.
No. Candidates, papers, approved text and audio are stored on the school's own server. Page images are processed on that same machine. The only case where anything leaves your network is if you deliberately enable the optional Azure cloud voice, in which case segment text is sent to Microsoft to be turned into audio.
Once the exam starts the app runs borderless full screen and topmost, hides admin navigation, blocks the usual close and minimise commands, and logs any attempt to leave the exam screen. It is designed for a supervised room — the invigilator remains the final control, and stronger Windows kiosk lockdown can be layered on if your district requires it.
Audio is cached before the sitting and the client tolerates a temporary network drop without ending an active session. We ask schools to put the server on a UPS and to disable sleep and Windows Update during the exam period — those two items are on the deployment checklist for a reason.
It is licensed per school rather than per learner seat, and the figure depends on your grades, learner numbers and whether you want on-site setup. Send us those details and we will put a written quote together.
Yes. We run structured pilots: a school uses Exam Reader for real mock exams across a full term, with agreed success criteria and a clear off-ramp if it is not working. Get in touch if you would like the pilot outline.
Bring Exam Reader to your school
Tell us your grades, learner numbers and languages and we will put together a quote and a demo date that suits your term calendar.