Sovereign Design Engine

I have projects that need to become physical objects. Photogeography is a photography archive that should exist as 11x17 tabloid posters. Versagrams is a song lyric series that should exist as 8.5x11 prints with per-song display typography and AI-generated imagery. The Deep Cuts, a book on DJ methodology and active listening, needs 6x9 bound spreads.

Making products feels integral to me as a human. The physical artifact is the point.

The production tool for this is InDesign. I don’t own it. And even when I did, the workflow was the same problem I keep hitting: the code lives in one environment, the design tool lives on another machine, and every time I need to produce something I leave the place where the content was built, open a different application on a different screen, and manually lay out what I already have structured as data.

Proprietary files, no version control, no scripting, every layout variant a manual file. I prefer to keep out of Adobe as much as possible.

So I built a headless pipeline. Source images and JSON manifests go in. Print-ready documents come out. Templates define layout as HTML and CSS. Manifests carry the data: a photograph’s EXIF data becomes coordinates, lens, timestamp, aspect ratio. A Genius API call becomes lyrics split across four columns. A markdown file becomes chapter content distributed across verso and recto pages. Paged.js polyfills CSS Paged Media into paginated documents with crop marks and bleeds. Vite serves the templates with hot reload for rapid iteration.

The core design constraint: every CSS value resolves to a fixed measurement. Inches for layout, points for type. Print production doesn’t tolerate ambiguity. The type scale is phi-derived (base 7pt, ratio 1.618). The spacing scale follows the same ratio from a quarter-inch base. Templates inherit both, so visual relationships stay consistent across different page sizes.

Three template types run through the engine now. Photogeography (part of Visual Studies) produces 11x17 tabloid posters with EXIF-driven metadata: GPS coordinates, camera, date, aspect ratio classification. Versagrams (also in Visual Studies) produces 8.5x11 posters with per-song display fonts, AI-generated imagery, and lyrics split into balanced columns. The Deep Cuts (a book in progress on DJ methodology) produces 6x9 book spreads with facing pages and margin notes.

Each template reads a JSON manifest. Each manifest can override CSS custom properties for per-piece adjustments. Every design is a text file in Git. Every build is deterministic: same manifest, same template, same output. The pipeline runs entirely offline, on my own hardware, with no subscription and no cloud dependency.