# autopresenter

> A local-first, agent-native studio that turns one Markdown composition and one machine-readable design system into on-brand artifacts: a narrated demo video (Cloudflare Workers AI Aura-2 voiceover), a self-contained HTML slide deck, a GIF, and a landing page. Slides are one projection of a shared brand IR, not the product.

Key facts:
- One source, many targets: `render <deck.md> --target slides|video` from a single Markdown composition.
- Slides → a single dependency-free HTML file (deep-linkable, presenter mode, Cmd-P → PDF). No Vite/Vue/Chromium.
- Video → SVG frames (rsvg-convert, no browser) narrated by Deepgram Aura-2, muxed with ffmpeg; pitch-preserving `--duration` time-fit; layout guards.
- DesignSystem is Markdown + YAML frontmatter: typed tokens + verbal config in frontmatter, messaging/copy in the body — shared parser and patch protocol with the brief and composition.
- Agent-native: every capability is a headless CLI command with JSON output and MCP parity.

## Docs
- [Documentation](https://autopresenter.openhackers.club/docs): composition format, design system, CLI, video pipeline, architecture.
- [PRD](https://github.com/OpenHackersClub/autopresenter/blob/main/PRD.md): full product spec, pillars, variants, milestones.
- [Composition format](https://github.com/OpenHackersClub/autopresenter/blob/main/specs/composition-format.md)
- [Design system](https://github.com/OpenHackersClub/autopresenter/blob/main/specs/design-system.md)
- [Why not Slidev](https://github.com/OpenHackersClub/autopresenter/blob/main/slidev.md): the cross-artifact-IR argument.

## Source
- [GitHub repository](https://github.com/OpenHackersClub/autopresenter)