Open your project, then either drag a file onto the Write studio's empty state or click "Upload manuscript" on the project's home tab. The file is parsed in the background; you'll see a progress indicator and can keep working while it runs.

When the import finishes the manuscript lands in the Write studio split into chapters (see Chapter detection). Every other studio that reads the manuscript, Edit, Pitch's synopsis builder, Sell's back cover, Market's comp tool, refreshes automatically.

Drop in any of: .docx (Word), .epub (the format every other writing platform exports), .pdf (text or rendered), and .txt. The cap is 20 MB per file, plenty for a 200,000-word novel even with images.

Other formats: Scrivener .scriv exports cleanly to .docx; export from Scrivener and upload the .docx. Google Docs: File → Download → .docx. Pages: File → Export to → Word. Markdown isn't directly supported yet but converts trivially via Pandoc.

Chapter detection runs automatically on upload. It looks for heading styles first (Heading 1 or Heading 2 in .docx; spine items in .epub; bookmarked outlines in .pdf), then falls back to text patterns like "Chapter 1" or roman numerals followed by a blank line.

If the result isn't right, open the chapter list and use "Re-split", you can choose between heading-style, text-pattern, or "every page break." You can also drag chapter boundaries by hand or merge two chapters into one. Splits are project-local and don't change the original file you uploaded.

For designed or already-typeset books, graphic novels, illustrated children's books, poetry chapbooks, a print proof you want to leave alone, flip on "Keep PDF as source" before uploading. The original PDF stays the canonical document and the studios that need text read it through OCR rather than re-flowing.

This means the Design studio is not available for that project (the layout is already done), but everything else works: the Pitch studio can still draft a query against the prose, the cover designer can iterate on a new cover, and the comp set tool reads the text as normal.

Standard OCR uses a fast extractive engine for clean, machine-set PDFs, included on every tier. If the upload looks scanned, hand-written, or visually complex, WriteLoom falls back to Claude vision OCR, which reads page images and produces structured text.

Vision OCR is significantly higher quality on hard inputs but takes longer; it runs in the background and updates each chapter as pages complete. Vision OCR is included on Loom and Tapestry; Spool gets the fast OCR pass only.

Every re-upload archives the previous version rather than replacing it. Open Manuscript → Versions to see the timeline, date, size, source filename, and to restore a previous version into the live editor.

If you want a clean working copy for a major revision, "Fork from version" creates a sibling project with the same plan notes, characters, and media but a fresh manuscript history. Useful for "what if I cut the second act" experiments without losing the original.