OCR → SVG stack
For any grimoire: recreate diagrams as editable SVG, insert them in the reader, and sell as digital asset pack / ebook / printed edition. Heptameron is the first full pipeline.
Production (use now)
| Tool | Role |
|---|---|
| VTracer | scan → vector facsimile |
| Surya OCR | historical printed text |
| Marker | PDF → structured blocks |
| OCRmyPDF + Tesseract | page text layer |
| figmark | docs + figures → markdown |
| SVGO + resvg | clean + deterministic render |
| Typst | print PDF from plates |
Semantic diagram recreation (moat)
| Tool | Role |
|---|---|
| Back2Struct (frontier) | diagram → object-level editable SVG |
| dots.mocr (MOCR) | multimodal OCR + graphics → SVG |
| StarVector | icons/technical diagrams |
| VFig | scientific figures → SVG |
| VIVID | text-heavy images → SVG with OCR |
| vectrify | raster → SVG via LLM search |
Grimoirer pipeline
- Scan → preprocess (dewarp / deskew / binarize)
- Text: Surya or Marker → blocks (reader text layer)
- Figure crop → VTracer facsimile (Source plate)
- Semantic SVG: Back2Struct / dots.mocr (editable asset)
- Procedural diagram.spec (ENDGAME moat)
- Insert into grimoire + sell pack / ebook / print
- Extract rituals/spells → astrology graph
Full notes: ochemapp/docs/OCR-SVG-STACK.md · Heptameron edition: /editions/heptameron · Asset pack: /downloads/heptameron/