Static, browser-based slide authoring for GitHub Pages.

This project is building a Markdown-driven presentation editor, compiler, and runtime that stays compatible with static hosting. The editor runs entirely in the browser, stores decks locally, and exports portable HTML presentations.

The authoring goal is closer to "Google Slides, but with Markdown" than a simple Markdown-to-slides converter. That includes responsive editing, deck-level review, safe experimentation, and eventually AI-assisted feedback that helps improve presentations without taking control away from the author.

This is intended for real production presentations. It is not a training product or a throwaway demo. The editor and output should help authors present with credibility, clarity, and accessible, trustworthy HTML.

• Markdown editor with live slide preview
• Front matter, optional generated title and closing slides, and slide splitting
• Note:, Resources:, and Script: support for presenter-facing context
• Audience view and presenter view with same-origin sync
• Slide outline and presenter timing support
• Progressive disclosure for lists
• Accessible LaTeX math (inline \( \) / $ $, display \[ \] / $$ $$) typeset with MathJax 4 — see docs/latex-math-formulas.md
• Fixed slide geometry with auto-fit body text and compact density warnings
• Better small-screen support for editor and audience view
• IndexedDB-backed local storage
• Service-worker-cached app shell for offline editing after first successful load
• In-editor AI prompt generator for briefing Ollama or another LLM
• Source import plus bundled ZIP export for Markdown, JSON, HTML, ODP, and MHTML
• Portable HTML snapshot export
• Print / Save PDF workflow with print‑friendly slide pages
• Optional caption/transcript support: the app can use a local whisper.cpp process or a configured transcript service, but Whisper is not bundled and not a git sub‑module; it must be run separately by the user.
• Built‑in theme presets plus optional external CSS override
• Accessible light/dark mode that respects system preference and supports manual override
• Presenter panel layout controls and countdown timer
• Accessibility checks in code for headings, links, image alt text, and notes
• Repo-level guidance for accessibility and coding agents

npm install
npm test
python3 -m http.server 4173

Then open:

• http://localhost:4173/
• http://localhost:4173/present/
• http://localhost:4173/presenter/

• AUTHORING.md: guide for slide authors — source format, front matter, theming, export, presenter controls, and optional captions
• AI_AUTHORING.md: compact authoring contract for external AI systems generating decks — strict syntax rules, the complete directive reference, and validating decks; see examples/ for canonical, validator-clean decks
• ACCESSIBILITY.md: accessibility posture, validation expectations, and contributor guidance
• AGENTS.md: repository instructions for coding agents and automation tools
• FEATURES.md: technical inventory of the implemented and maintained feature set
• LICENSE: AGPLv3 license for the project
• COMPARISON.md: feature comparison against other Markdown presentation tools
• SBOM.md: software bill of materials with technology versions and licenses used
• STYLES.md: writing, design, and styling standards for the app and repository docs
• .agents/skills/: repo-local GitHub Copilot accessibility skills for planning, testing, and reviewing UI interactions
• .github/copilot-instructions.md: onboarding notes for GitHub Copilot coding agent
• TODO.md: future roadmap and integration ideas
• docs/accessibility-checklist.md: project accessibility checklist
• docs/ai-authoring-workflow.md: recommended workflow and prompt template for AI-assisted deck drafting
• docs/cloudflare-sync-plan.md: optional Cloudflare Worker + D1 plan for authenticated online save, load, and comments
• docs/editor-vision.md: longer-term product direction for the editor and runtime
• docs/latex-math-formulas.md: writing LaTeX math formulas in slides, with copy/paste examples
• docs/layout-syntax.md: layout directives for centered content, columns, media blocks, callouts, and quotes
• docs/live-captioning-guide.md: event room setup, browser-native and whisper.cpp captioning, volunteer guidance, known issues, and privacy
• docs/speech-to-text-architecture.md: full technical description of the browser STT pipeline (Web Speech API, whisper.cpp bridge, transcript cleanup prompt engine, BroadcastChannel audience sync)
• docs/manual-a11y-testing.md: Sa11y-assisted and manual accessibility review workflow
• docs/resources.md: accessibility references and source material

Disclosure of AI use is important to this project.

• OpenAI Codex, a GPT-5-based coding agent in the Codex desktop environment, was used during repository setup and early development. It was used to scaffold the initial static application, write and revise project documentation, add tests, update planning and guidance files, and diagnose and revise iframe fallback and offline-cache behavior.
• opencode (assistant mode) was used for a 2026 code and accessibility review pass. It fixed stale BDD export and math expectations, corrected the editor first-slide control state, added keyboard guards so presentation shortcuts no longer hijack focused buttons, selects, or checkboxes, and scoped presenter timer/caption redraws to their own UI nodes instead of rebuilding slides every second.

• No AI is required to run the program by default.
• The current application runs as a static browser app with no required runtime dependency on an LLM or remote AI service.
• Optional speech-to-text can be enabled for presentations when a local whisper.cpp process or a compatible transcript service is available.

• Browser-based AI is not currently enabled in this application.
• No part of the current shipped editor or presentation runtime executes an in-browser LLM.
• Caption UI stays hidden unless a transcript source is actually available.

• Optional AI-assisted review and editing may be added later.
• If implemented, AI use should be disclosed clearly, kept optional, and separated from the static baseline experience.

• / editor
• /present audience presentation
• /presenter presenter view

The repository includes 404.html so GitHub Pages can redirect deep links back into the single-page app.

• index.html loads the app shell
• src/main.js routes between editor, audience, and presenter views
• src/modules/parser.js parses front matter, slides, and presenter support sections
• src/modules/markdown.js renders the current supported Markdown subset
• src/modules/a11y.js contains accessibility lint rules
• src/modules/captions.js handles caption-source detection, transcript parsing, and transcript polling
• src/modules/export.js generates the standalone snapshot HTML
• src/modules/presentation-state.js manages reveal-step navigation and duration metadata
• src/modules/presenter-timer.js manages presenter countdown timer state
• src/modules/slide-layout.js manages slide dimensions and body-text fitting
• src/modules/theme.js applies built-in themes and optional external stylesheet overrides
• src/modules/color-mode.js manages accessible light/dark mode
• src/modules/storage.js handles IndexedDB-first persistence
• styles/app.css contains the UI and presentation styling

Accessibility applies to both the editor UI and the generated slide output. The project targets WCAG 2.2 AA for both surfaces, covering accessible HTML, keyboard navigation, visible focus states, and reduced-motion support.

Live captions are a first-class accessibility feature. The presenter view supports three captioning modes — browser-native Web Speech API, local whisper.cpp, and external caption services — satisfying WCAG 2.2 SC 1.2.4 — Captions (Live). The caption panel stays hidden until a source is confirmed available, so the static baseline is never affected. See docs/live-captioning-guide.md for event setup and ACCESSIBILITY.md for the full conformance posture.

Repo-local GitHub Copilot skills are available for accessibility work, including a11y-planner, a11y-critic, a11y-test, a11y-workflow, and perspective-audit. They can be used to probe button behavior such as the editor page's previous/next slide controls against the default deck.

Whisper-style speech-to-text remains optional. AI-related controls should only be exposed when local or API-backed AI is actually available. Static GitHub Pages mode should continue to work without any server or AI dependency.

Longer term, the editor should support optional LLM-assisted slide review and revision workflows, including local-first integrations such as Ollama, comment-style suggestions, and strong undo/redo support for all applied changes.

The project should be reusable by others, including teams such as CivicActions and W3C, but it should not depend on their adoption to justify product quality. The baseline should stand on its own.

The accessibility approach should actively build on strong public guidance, especially Intopia's presentation accessibility recommendations, while also promoting resources from Inklusiv and related accessibility communities.