About This Project

Mission

The GLOW Accessibility Toolkit helps authors, publishers, and organizations produce documents that comply with the American Council of the Blind Large Print Guidelines and WCAG 2.2 Level AA. The toolkit audits, fixes, and templates Word, Excel, PowerPoint, Markdown, PDF, and ePub documents for accessibility.

GLOW stands for Guided Layout & Output Workflow -- a guided, confidence-building accessibility experience built for real publishing workflows.

Organizations

American Council of the Blind (ACB): The nation's leading membership organization of blind and visually impaired people. ACB publishes the Large Print Guidelines through its Board of Publications, which define minimum standards for font, size, spacing, emphasis, and layout in large print materials.
Blind Information Technology Solutions (BITS): A special interest affiliate of the American Council of the Blind. BITS members are professionals in the technology field who are blind or visually impaired. BITS sponsors this open source project to advance document accessibility tooling.
W3C Web Accessibility Initiative (WAI): Develops the Web Content Accessibility Guidelines (WCAG), which this toolkit uses for digital content requirements including contrast, reflow, and line spacing.
DAISY Consortium: An international association of libraries and organizations serving people with print disabilities. DAISY develops the Ace accessibility checker, the Accessibility Metadata Viewer, and the Accessible Publishing Knowledge Base used by this toolkit for ePub accessibility.

Standards and References

ACB Large Print Guidelines -- revised May 6, 2025. The canonical specification for large print formatting: Arial 18pt body, 22pt headings, 20pt subheadings, flush left alignment, underline-only emphasis, no italic.
WCAG 2.2 -- Web Content Accessibility Guidelines Level AA. Applied to digital output: 4.5:1 contrast ratio, 400% zoom reflow, 1.5x line-height.
Microsoft Accessibility Checker -- rules for Word, Excel, and PowerPoint document accessibility (alt text, table headers, reading order, hyperlink text).
PDF/UA (ISO 14289-1) -- the international standard for accessible PDF documents. Used for PDF audit rules.
EPUB Accessibility 1.1 -- W3C specification for accessible EPUB publications. Covers discovery metadata, access modes, navigation, reading order, and conformance to WCAG.
DAISY Accessible Publishing Knowledge Base -- the DAISY Consortium's comprehensive reference for accessible publishing practices, covering headings, images, tables, MathML, page navigation, and metadata.

Open Source Dependencies

This project is built with open source software. We gratefully acknowledge the following projects:

Core Libraries

Python libraries used by the GLOW Accessibility Toolkit
Library	Version	License	Purpose
Python	3.13+	PSF	Runtime language
Flask	3.1+	BSD-3-Clause	Web application framework
Flask-WTF	1.2+	BSD-3-Clause	CSRF protection and form handling
Flask-Limiter	3.5+	MIT	Rate limiting for abuse prevention
Gunicorn	23.0+	MIT	Production WSGI server
python-docx	1.1+	MIT	Read, audit, and modify Word (.docx) documents
Mammoth	1.8+	BSD-2-Clause	Convert Word documents to clean HTML
openpyxl	3.1+	MIT	Read and audit Excel (.xlsx) workbooks
python-pptx	1.0+	MIT	Read and audit PowerPoint (.pptx) presentations
PyMuPDF (fitz)	1.24+	AGPL-3.0	Read and audit PDF documents
MarkItDown	0.1+	MIT	Convert documents and audio to Markdown (Microsoft)
DAISY Ace	1.3+	MIT	EPUB accessibility checker -- 100+ axe-core rules + EPUB metadata validation (DAISY Consortium)
DAISY a11y-meta-viewer	2.0	CC BY-NC-SA 4.0	Reference implementation of the W3C Accessibility Metadata Display Guide 2.0 -- generates human-readable accessibility statements from EPUB metadata (DAISY Consortium)
Pandoc	3.1+	GPL-2.0	Universal document converter -- transforms Markdown, Word, reStructuredText, OpenDocument, Rich Text, and ePub into accessible HTML, Word (.docx), EPUB 3, and PDF (via WeasyPrint) with embedded ACB Large Print formatting (John MacFarlane, University of California, Berkeley)
WeasyPrint	62.0+	BSD-3-Clause	CSS-based PDF rendering engine -- converts Pandoc HTML output to accessible PDF with ACB BOP print formatting (Arial 18pt, 1.15 line spacing, 1-inch margins, @page rules). Uses Liberation Sans font (metrically identical to Arial) for consistent print typography (CourtBouillon)
DAISY Pipeline 2	1.15+	LGPL-3.0	Advanced document conversion engine -- Word to EPUB 3, HTML to EPUB 3, EPUB to DAISY 2.02, EPUB to DAISY 3/DTBook. Used for producing accessible e-books and talking book formats (DAISY Consortium)

Optional and Build Dependencies

Optional libraries used for GUI, testing, and packaging
Library	Version	License	Purpose
wxPython	4.2+	wxWindows (LGPL-like)	Desktop GUI (wizard interface)
pytest	8.0+	MIT	Test framework
PyInstaller	6.0+	GPL-2.0 (with exception)	Windows desktop executable packaging
Inno Setup	6.0+	Inno Setup License (free)	Windows installer (.exe) creation

Infrastructure

Infrastructure and deployment components
Component	License	Purpose
Docker	Apache 2.0	Container runtime for web deployment
Caddy	Apache 2.0	Reverse proxy with automatic HTTPS
OpenJDK (JRE)	GPL-2.0 + Classpath	Java runtime required by DAISY Pipeline 2

Project Information

Source code: github.com/accesswatch/acb-large-print-toolkit
Changelog: View the complete project history
License: MIT License
Author: Jeff Bishop, President of BITS
Current release: 1.0.0
Desktop tool version: 1.0.0
Web application version: 1.0.0
Python version: 3.13+

Stress Testing

The toolkit includes a reproducible synthetic Word-document stress harness for the heading-detection and fixing pipeline. The current corpus covers 1000 heading scenarios across 1000 generated documents.

These generated documents deliberately mix ACB-conformant content with failure patterns such as centered titles, justified paragraphs, hanging indents, faux headings, signature blocks, copied web formatting, and body-text emphasis drift. Each failure that proves useful is folded back into the audit rules, fixer behavior, or online help so the product learns from the corpus instead of letting that knowledge stay trapped in tests.

Detection coverage: 10 scenario families spanning paste workflows, agendas, newsletters, policy manuals, flyers, and appendices.
Fixer coverage: Every generated repair document is validated for flush-left alignment and exact paragraph-indent repair after fixing.
ACB strictness: Whenever the harness generates an ACB-guideline branch, all text is left aligned by design.

Validation transparency: For the current 1,000-document scale, maintainers measure four things: full-suite test health, direct detector-vs-ground-truth accuracy, denser randomized accuracy, and full fix-then-audit ACB compliance. The current measured runs were clean.

What we learned: The biggest risks were not crashes. They were subtle document-policy gaps, especially short false positives and correctly detected headings that still needed hierarchy or text cleanup after repair.

How the platform changed: We strengthened false-positive handling, widened random scenario coverage, and added heading normalization so converted headings no longer leave behind ALL CAPS text or skipped levels.

What the documents represented: The generated files stood in for the kinds of Word documents people really upload: agendas, newsletters, policy manuals, legal outlines, training handouts, appendices, email-style content, and plain-text paste with no real heading styles. The goal was to test realistic document behavior, not just isolated heading samples.

Acknowledgments

This project would not be possible without the tireless advocacy of the American Council of the Blind and the volunteers who authored and maintain the ACB Large Print Guidelines. Special thanks to the ACB Board of Publications for establishing clear, enforceable standards that make printed and digital materials accessible to people who are blind or have low vision.

We gratefully acknowledge the DAISY Consortium for their leadership in accessible publishing. This toolkit relies on three DAISY projects: DAISY Ace for EPUB accessibility validation, the DAISY Accessibility Metadata Viewer for human-readable display of EPUB accessibility metadata per the W3C Accessibility Metadata Display Guide 2.0, and DAISY Pipeline 2 for advanced document-to-EPUB and EPUB-to-DAISY conversions that produce accessible publications for e-readers, refreshable braille displays, and DAISY players.

The HTML conversion engine is powered by Pandoc, the universal document converter created by John MacFarlane. Pandoc transforms Markdown, Word, reStructuredText, OpenDocument, Rich Text, and ePub files into accessible HTML with ACB Large Print styling.

Thanks also to the BITS community for testing, feedback, and ongoing development support, and to the open source maintainers whose libraries make this toolkit possible.