Flagship exemplar

Equation‑dense STEM, made truly accessible.

A scanned electromagnetics chapter has its mathematics trapped in images — a screen reader gets noise, not "the complex propagation constant equals omega root mu‑epsilon." Here is the same chapter as a certified accessible HTML + MathML edition: the math spoken and explorable by a screen reader — built from the image‑only PDF, and proven 100% faithful to the source.

Open the accessible edition → See the QA & provenance (signed)

Side-by-side comparison. Left, labeled BEFORE, image-only PDF: the four vector Maxwell equations 2.79 to 2.82 as a scanned picture a screen reader cannot read. Right, labeled AFTER, accessible edition: the same four equations as live MathML that a screen reader speaks aloud. — The same four equations (Maxwell's equations in a lossy medium, 2.79–2.82): trapped as pixels in the scanned PDF (left), and rebuilt as live, spoken MathML (right).

What the exemplar proves

0extractable characters in the source — a genuine image‑only PDF

100%fidelity: 30/30 equations character‑exact vs. the author's LaTeX

0WCAG 2.1 AA violations (axe); contrast measured, all pass

109expressions rendered as real MathML — spoken & explorable

Signednamed author fidelity review & sign‑off

Hear the math

Audio rendering — the equations read aloud in a screen‑reader‑style voice (neural TTS), so you can hear what the accessible math sounds like. This is an illustrative narration, not a live screen‑reader capture.

Coming: a live recording of a blind screen‑reader user reading the edition and exploring an equation term by term — the strongest proof of real‑world accessibility, validated by a native screen‑reader user.

How it was made

From the image‑only PDF, not the source. The pages were rasterized to images (no text layer) and the mathematics transcribed from those images into real MathML — the same source‑less path a real scanned client text requires.
Verified against ground truth. The image‑derived transcription was then checked against the author's original LaTeX and matched 100%, character‑for‑character.
Documented, human‑reviewed, gated. Every step produces a checkable output; ambiguities were surfaced for the author, not silently guessed; the author signed off. It is not a black box.
Spoken by real assistive tech. The math is real MathML rendered with MathJax and a speech engine, so NVDA + MathCAT (Windows) and VoiceOver (macOS) read and navigate every equation.

The material is the author's own work, reproduced with permission and freely shareable. This edition describes a documented, human‑reviewed conversion process and makes no compliance guarantee.