The guide
Ten minutes here saves an evening later. The engine reads your manuscript's structure — mark it clearly, and your book sets right the first time.
The one rule
TimePress doesn't guess at a design — it typesets your book from its structure: chapters, sections, paragraphs. Everything downstream (right-hand chapter openings, running heads, the contents page, the folios) hangs off those marks. A manuscript with clear chapter headings sets beautifully; a manuscript where the engine has to guess produces a wall of warnings.
The good news: "clear" means one thing only — chapter titles carry a real heading style. That's the whole trick.
File formats
In order — from "just works" to "last resort"
The gold standard. Style each chapter title as Heading 1 (sections within a chapter as Heading 2) and the engine reads your structure with certainty — no guessing at all. In Google Docs: File → Download → Microsoft Word (.docx).
# chapter headingsJust as reliable as .docx. Put # before each chapter title, ## before sections, and leave a blank line between paragraphs.
Works, but plain text carries no heading marks, so the engine has to infer where chapters start. Usually decent; never as sure as a styled document. If you have five minutes, add # marks and save it as .md instead.
The engine has to scrape the text back out — with the old edition's page numbers and running headers mixed in — and guess the structure from the debris. Expect noisy findings. A scanned, image-only PDF can't be read at all; run OCR or find the text another way.
The checklist
Do these once, and the engine does everything else
Reviving an old book
Setting a public-domain classic — or re-setting a family book that only survives in print — is exactly what TimePress is for. Just don't feed it the old edition's PDF: that's a finished layout, and the engine will fight it.
Nearly every public-domain book is on Project Gutenberg as clean plain text. Download that, put a # before each chapter heading (ten minutes with find-and-replace), save it as .md, and the engine will set a genuinely fresh edition rather than exhuming a hundred-year-old one. For a book that exists only on paper, scan it, run OCR, and tidy the text the same way.
After you upload
The engine tags your structure, typesets the pages, then a deterministic checker reads every page of the result — widows, orphans, chapter openings, running heads — and tells you honestly what it found. Findings point at pages; the Spread viewer shows you each one in context.
TimePress never edits your words, so fixes happen at the source: adjust the manuscript, upload it again, and build. Each build is set fresh from the file you give it — which is also why the same manuscript always produces the same book.
Access is by entry code while we're in early release.