Convert CHM to TXT

How it works

The job has two deliberately separate transformations. extract_chmLib extracts HTML from the CHM into temporary storage, and then the selected TXT output is produced from that HTML by removing markup and writing readable text. This makes the route useful for plain-text workflows without pretending that a TXT file can retain the original help viewer, stylesheet, image directory, or contents pane.

HTML contains clues about structure, but plain text has fewer ways to express them. A heading can become a line, a list can become indented text, and a paragraph can remain separate when its markup is conventional. A table may collapse into a sequence of cells, and a code sample can lose the spacing that distinguished one column from another. The exact result depends on how the topic was authored, not just on the CHM extension.

Why use plain text

TXT is easy to search, diff, archive, and pass between tools. It is a sensible target for an internal index, a command-line review, a quote extraction task, or an accessibility pipeline that wants to apply its own presentation. Removing CSS can also remove visual noise from a documentation set whose main value is the wording of warnings and commands.

That simplicity has a cost. Images, diagrams, colors, fonts, scripts, and most links have nowhere to go. The conversion does not promise that CSS or assets will be carried along, and it does not merge an entire help project into a perfectly labeled text book. If topic identity matters, add filenames or headings through a local editorial process and keep the original package beside the generated text.

Check the content that can disappear

Search for a warning, a command with punctuation, an accented name, and a phrase from the last topic. Inspect a table and a code listing separately. If the manual teaches a task through screenshots, the TXT output is incomplete by design; attach an authorized text description or refer readers back to the CHM. A successful download proves that text was written, not that every visual instruction or navigation relationship survived.

What a CHM is, and why plain text loses so much of it

Compiled HTML Help is a Microsoft format from 1997: a compressed archive of HTML pages plus an index, a table of contents, and a full-text search database, all built into one .chm file. It was how Windows software shipped its manuals for about fifteen years, and a great deal of documentation exists nowhere else.

The format is a small website with a navigation apparatus around it. Extracting text keeps the pages and discards the apparatus, and the apparatus was doing real work.

The table of contents is the first casualty. A CHM's structure is usually hierarchical — chapters containing topics containing sub-topics — and that hierarchy lives in a separate index file rather than in the pages. Flattened to text, the topics arrive in sequence with no indication of what was nested under what.

Cross-references go next. Help files are heavily interlinked, and "see the section on printing" was a link. In plain text it is a sentence pointing at nothing, and there is no way to tell from the output what it pointed to.

The pop-up definitions and context-sensitive topics that these files used extensively simply have no text equivalent at all.

Take the extraction when you need the words — searching for a phrase, feeding documentation to a text tool, preserving content from a file whose reader no longer runs. Where the structure matters, decompiling the CHM to its original HTML keeps it.

Old files, old encodings

CHM files predate the general adoption of Unicode, and the ones you are likely to encounter were written when Windows code pages were how non-English text got stored.

A file authored in Western Europe is probably Windows-1252; one from Central Europe, Windows-1250; Cyrillic, Windows-1251; Japanese, Shift-JIS. The encoding is sometimes declared inside the HTML and sometimes not, and when it is not, a converter has to guess.

The symptom of a wrong guess is recognizable. Accented characters arrive as pairs of odd symbols — é where é belongs — or as question marks and empty boxes. Pure ASCII text is unaffected, so an English manual usually comes through perfectly and a German one may not.

Check for it deliberately rather than waiting to notice: search the output for a word you know contains an accent, or for a currency symbol. Both are quick and both fail visibly.

If the text is mangled, the content is still recoverable — the bytes are intact and only their interpretation is wrong. A text editor that lets you reopen a file with a specified encoding will usually fix it in one step.