DunneWrite · Tutorial

Import a PDF novel

Convert a PDF to clean text, remove the parts that would mislead the AI, and ingest it.

Available20 min· Intermediate· DunneWrite 0.22.0

DunneWrite reads text, not PDF. If you pick a .pdf, the app stops and shows you how to convert it instead of importing it, because a PDF read as text comes out as nonsense. In this tutorial you'll convert a PDF novel, clean it up, and ingest it.

You'll need#

A PDF of the book, whose text can be selected (not a scan of page images).
The free pdftotext tool, or any other PDF-to-text or PDF-to-Markdown converter:

Platform	Where `pdftotext` comes from
Windows	It ships with Git for Windows. Open Git Bash and it is already there.
Linux	The `poppler-utils` package
macOS	`brew install poppler`

A text editor.
A provider connected in DunneWrite. See Connect an AI provider.

Convert the PDF#

Open a terminal (on Windows, Git Bash) in the folder that holds the PDF.

Run:

pdftotext -layout "your-book.pdf" "your-book.txt"

Open your-book.txt and check that the prose reads normally.

Keep -layout. Without it, pages with any column structure interleave their lines and the prose comes out shuffled. Conversion is fast: a 1,200-page novel converts in a few seconds.

Clean it up#

Conversion faithfully reproduces two things that belong to the printed page rather than to the book. Fix both before importing.

Rejoin words split across a line break. Typesetting hyphenates words at the right margin, so mur- at the end of one line and mured at the start of the next should become murmured. Rejoin only when the next line starts with a lower-case letter, so that real hyphenated names such as Barrow-Downs survive. A long novel can have hundreds of these.
Cut everything that is not the story. Delete the front matter, the table of contents, the appendices and, above all, the index.

Ingest it#

Optionally, save the cleaned file as Markdown (.md); Markdown is the preferred format.
In DunneWrite, open Ingestion Review and press Ingest a file — see the cost first….
Pick the cleaned file, and answer honestly when asked whose writing it is and what kind of material it is.
Read the cost screen, then start the run. A whole novel is a large run; consider trying a few chapters first.

What you've learned#

DunneWrite takes text, and pdftotext -layout is a reliable way to get it from a PDF.
Rejoin line-break hyphens and remove the index before importing.
Try a slice of a large book before committing the whole thing.

Next#

Ingest your notes, for reviewing what was found
Place characters on the Globe

Something unclear or out of date? Open a support ticket.

DunneWrite · Tutorial

Import a PDF novel

Convert a PDF to clean text, remove the parts that would mislead the AI, and ingest it.

Available20 min· Intermediate· DunneWrite 0.22.0

You'll need#

A PDF of the book, whose text can be selected (not a scan of page images).
The free pdftotext tool, or any other PDF-to-text or PDF-to-Markdown converter:

Platform	Where `pdftotext` comes from
Windows	It ships with Git for Windows. Open Git Bash and it is already there.
Linux	The `poppler-utils` package
macOS	`brew install poppler`

A text editor.
A provider connected in DunneWrite. See Connect an AI provider.

Convert the PDF#

Open a terminal (on Windows, Git Bash) in the folder that holds the PDF.

Run:

pdftotext -layout "your-book.pdf" "your-book.txt"

Open your-book.txt and check that the prose reads normally.

Keep -layout. Without it, pages with any column structure interleave their lines and the prose comes out shuffled. Conversion is fast: a 1,200-page novel converts in a few seconds.

Clean it up#

Conversion faithfully reproduces two things that belong to the printed page rather than to the book. Fix both before importing.

Rejoin words split across a line break. Typesetting hyphenates words at the right margin, so mur- at the end of one line and mured at the start of the next should become murmured. Rejoin only when the next line starts with a lower-case letter, so that real hyphenated names such as Barrow-Downs survive. A long novel can have hundreds of these.
Cut everything that is not the story. Delete the front matter, the table of contents, the appendices and, above all, the index.

Ingest it#

Optionally, save the cleaned file as Markdown (.md); Markdown is the preferred format.
In DunneWrite, open Ingestion Review and press Ingest a file — see the cost first….
Pick the cleaned file, and answer honestly when asked whose writing it is and what kind of material it is.
Read the cost screen, then start the run. A whole novel is a large run; consider trying a few chapters first.

What you've learned#

DunneWrite takes text, and pdftotext -layout is a reliable way to get it from a PDF.
Rejoin line-break hyphens and remove the index before importing.
Try a slice of a large book before committing the whole thing.

Next#

Ingest your notes, for reviewing what was found
Place characters on the Globe

Something unclear or out of date? Open a support ticket.