How to Extract Plain Text from a PDF (Selectable + Scanned, In Browser)

You have a PDF. You need the words, not the layout. To feed into an LLM. To search with grep. To quote in another document. To pipe through wc -w for a word count. To put into a translation tool. To version-control with git.

PDF copy-paste mostly works but mangles line breaks, drops hyphenated words, and inserts random spaces. A real text extractor handles all that.

Why local extraction makes sense

PDF text often holds the actual content of confidential documents — contracts, reports, drafts, financials. If you upload to extract, you’ve shipped your sensitive content to a third party for the price of saving 30 seconds. The browser extractor uses pdf.js (Mozilla’s official PDF library) to walk the text stream, applies reading-order heuristics, and outputs .txt — all locally.

Step-by-step

1. Open PDF to Text.
2. Drop your PDF.
3. Settings:
   • Preserve layout (default off): off = cleaner prose flow; on = keeps line breaks and indentation matching the PDF
   • Include page markers: inserts --- Page N --- between pages
   • Pages: all, or a specific range
4. Click Extract. Preview shows in a few seconds.
5. Download .txt or copy from the preview pane.

Born-digital vs scanned PDFs

If you can select text in your PDF reader (Acrobat, Preview, your browser), the PDF is born-digital. Text extracts directly. Single-page PDF: under a second. 500-page PDF: 5-10 seconds.

If the text is “trapped” — you can see it but can’t select it — the PDF is scanned/image-only. You need OCR first:

1. OCR PDF — adds a hidden text layer to the scanned PDF
2. Then back to PDF to Text — extracts that layer

Both tools run in browser. OCR uses tesseract.js (~98% accuracy on clean 300 DPI prints).

Use cases this is good for

• Feed to ChatGPT / Claude / Gemini — paste cleanly, no markdown noise from formatting
• Search across many PDFs — extract to .txt, then grep your folder
• Version control drafts — .txt in git diffs readably; .pdf doesn’t
• Translation — most translation tools work better on plain text than on PDF
• Word count for billing/contracts — wc -w extracted.txt
• Building a private knowledge base — extract PDFs, index in your local search tool

What you lose

PDF to Text drops:

• Images (entirely)
• Font/color/size styling
• Table structure beyond tab-separated rows
• Hyperlinks (just the link text, not the URL)
• Form fields (the labels show, the input values don’t)

If you need structure preserved (headings, lists, tables), PDF to Word keeps a closer-to-original layout in a .docx you can edit. If you need the layout fully intact, keep it as PDF.

Tips for clean output

• Hyphenated line-end words: most PDFs hyphenate “informa-” / “tion” across lines. The extractor rejoins these by default. Toggle off if you’re processing the actual line-break locations.
• Page numbers in body text: turn on “remove repeating headers/footers” to drop boilerplate page numbers and document titles that appear on every page.
• Bullet markers: •, -, *, and similar are kept as-is. If you want to strip them, post-process the .txt with a one-line sed.

Privacy posture

Every step — PDF parsing, OCR if needed, text export — happens in your browser. The PDF is read into memory, processed, and the resulting text is generated locally. Network is only used to load the page itself. For sensitive documents (legal, medical, financial), this is the only correct architecture.

After extracting

The .txt is yours. Some common chains:

• Pipe to your LLM of choice for summarization / question answering
• Feed into Translate PDF if you actually want the PDF translated (it takes the source PDF, not the extracted .txt)
• Use Compare PDF if you need to diff two related PDFs (it has a built-in diff better than diff on extracted text)

Or just open in your editor and start writing.

Frequently asked questions

Related articles