PDF → TXT

What is PDF?

PDF is Adobe's fixed-layout document format. It records what to draw at which coordinate on a page rather than which paragraph a sentence belongs to. The layout is therefore identical everywhere, which suits contracts, papers and scans. The cost is that PDF is inherently hostile to reflow: line breaks are usually visual, extracted text often carries hard returns and scrambled column order, and scanned PDFs contain images only, with no text layer at all.

What is TXT?

TXT is plain text: the file holds characters and nothing else, with no fonts, sizes, colours, images or pagination. Line breaks are its only structure. Most Chinese web novels circulate as TXT precisely because it is small, opens anywhere and is easy to copy and update incrementally. The trade-offs are that it carries no chapter structure, cover or metadata, and that mismatched encodings (UTF-8, GBK, UTF-16) produce mojibake — the two most common TXT problems.

What to watch out for when converting PDF to TXT

Extracting from the source

Line breaks in PDF are mostly visual, not paragraph ends: one paragraph may arrive as five separate text blocks, each with its own hard return. Merging them back requires heuristics based on line width and sentence endings. A scanned PDF has no text layer at all, so extraction returns nothing until you run OCR.

Writing the target format

TXT stores characters only: every image, font, size, colour, table, footnote and hyperlink is gone, and headings degrade to ordinary lines. You must also choose an encoding (UTF-8 for Chinese, GBK if targeting older devices) — the wrong choice yields immediate mojibake.

Where the two formats interact

• Layout must be rebuilt: the paragraph, column and figure positions in the source were arranged for a fixed page and mean nothing once the target reflows. The converter can only re-stitch a text stream, so paragraphs may merge wrongly, columns may interleave and figures may land mid-sentence. Read the first few chapters afterwards, watching dialogue breaks and figure placement.
• The source is one full-page image per page while the target reflows: each image is inserted in sequence, and with no text layer you end up with an image stream. The result is usually huge and unsearchable — a worse reading experience than the original.
• Title, author, publisher, language and cover have nowhere to live in the target, so all metadata is lost. If the result goes back into a library, the filename is the only place left to record that information.
• The source has no chapter concept (comics and PDFs are page-based) while the target expects navigation. The converter can only split mechanically by page or fixed length, producing a semantically empty TOC. Ignore it if you do not need navigation, but do not expect real chapters to appear.
• Any conversion touching TXT must settle encoding: Chinese TXT files are commonly UTF-8, GBK, GB18030 or BIG5, with or without a BOM. Write UTF-8 without BOM for best compatibility, or GBK for legacy devices. A wrong choice shows up as wholesale mojibake, not as an error.
• Line endings need normalising too: CRLF on Windows, LF on Unix, and historically CR on classic Mac. Mixing them makes some readers render everything as a single line — if the whole book collapses into one paragraph, check line endings first.

Post-conversion checklist

1. Convert the first three chapters as a trial and check for lost text, interleaved lines or misplaced blocks before doing the whole book.
2. Check chapter splitting: the TOC entry count should match the real chapter count.
3. Jump to five random spots and search a character name to confirm the text is fully searchable.
4. Verify encoding and line endings: UTF-8 without BOM, line endings normalised to LF or CRLF.

Converting novels: what is different

Fiction is the most common and the most peculiar kind of e-book: hundreds or thousands of chapters, almost no figures, often nothing but a TXT source, and Chinese sources that frequently carry anti-piracy garbling. These are the points that bite hardest.

• A long novel can run to hundreds or thousands of chapters, so the table of contents decides whether the result is usable. Normalise chapter headings to one recognisable pattern first — say "Chapter N Title" alone on its own line — or the target will miss chapters, merge them, or collapse the whole book into one.
• Chinese web novels in TXT show three very different kinds of garbled text. (1) Encoding mismatch — reading UTF-8 as GBK; fixing the encoding restores it. (2) Anti-piracy substitution — the author deliberately replaced characters; it cannot be recovered, so find another source. (3) Full-width/half-width mixing or missing rare glyphs — a font problem. Diagnose which one you have; do not keep re-encoding the third case.
• Covers, character art and volume-opening illustrations are all dropped, and the library shows a blank cover. Serialised novels often open each volume with an image; afterwards only an empty line is left where it used to be.
• Volume markers ("Volume One", "Volume Two") usually degrade to ordinary headings, losing the volume/chapter hierarchy. If that matters, promote volume headings one level in an editor afterwards so they are not siblings of chapter titles.

Tools for converting PDF to TXT

Any of the tools below can handle this direction; pick one by your operating system and how many files you have. Read the previous section first — doing the steps out of order (converting a scanned file before running OCR, for instance) usually means starting over.

Common questions

Related conversions