Getting content out of a PDF: text, data, images and archives

Why PDFs are hard to convert back

Inside most PDFs, text is stored as characters placed at exact positions on a page. There's often no record that a group of lines forms a paragraph, that certain words are a heading, or that a grid of numbers is a table with rows and columns. A converter has to rebuild that structure from positions, which works well for simple documents and poorly for complex layouts. Scanned PDFs are harder still, because they contain no characters at all, only pictures of pages. That's why the most reliable path is always the original source file when it exists, and why VelloDoc's converters favor clean, honest output over imitations of the original layout. The format guide explains the underlying trade-offs.

Recovering the words

Three converters recover text for different destinations. PDF to Word produces an editable DOCX with the text of each page as paragraphs and page breaks preserved, ready to restyle, but it doesn't rebuild fonts, tables, images or columns. PDF to TXT writes all text to a UTF-8 file, the most portable form for quoting, searching and analysis. PDF to Markdown adds a heading for each page, a convenient starting point for wikis and note apps. If the output looks garbled even though the page reads normally, the PDF stores characters without a proper mapping to letters. Running OCR and converting again usually fixes it.

Recovering numbers and tables

PDF to Excel creates one worksheet per page and turns each line of text into a row. It splits cells wherever the text has a tab or a wide gap, since that's how many PDFs lay out columns. Simple, evenly spaced tables come across well. Merged cells, wrapped text and empty cells need tidying, because spacing is the only clue to structure. Values arrive as text, so convert them to numbers before calculating. If the data first came from a system export, asking for the CSV file is always more accurate than pulling it out of a PDF. And when you need a printable table, CSV to PDF does the reverse.

Pictures of pages or the images inside

Two different jobs are easy to confuse. Turning pages into images gives you a picture of each whole page. PDF to JPG, PDF to PNG and PDF to WebP convert every page at 120, 150 or 200 DPI and give you a ZIP. JPG suits photos, PNG suits text and diagrams, and WebP suits websites. Extract Images instead pulls out the photos and graphics stored inside the PDF, in their original format and resolution. It can be far better quality than a page render. For presenting pages, PDF to PowerPoint places each page on a slide sized to the page shape, as an image you can annotate but not edit.

A PDF as a web page

Long PDFs are awkward on phones and opaque to search engines. PDF to HTML turns a document into a single web page file. Pages with a text layer become real text that fits the screen and can be selected, searched and read aloud, with their images included. Scanned pages are added as page images, so nothing is lost. Exact layout, columns and fonts are simplified in favor of readability. For documentation sites that use Markdown, the Markdown converter is the better fit, and publishing the original PDF alongside the HTML keeps a printable version available for people who want one.

Scans need OCR first

Every text-based conversion depends on a text layer. A scanned or photographed PDF has none, so PDF to Word returns empty pages, PDF to Excel returns empty sheets and PDF to HTML falls back to page images. OCR PDF adds recognized text in any of 39 languages, after which every text converter works normally. Recognition isn't perfect, especially with small print, handwriting and complex layouts, so check names, numbers and dates in the converted result. Cleaning scans before recognition, with the optimization tools for straightening and blank-page removal, measurably improves accuracy, and what makes a PDF searchable explains why.

Answers from many PDFs at once

Sometimes you don't need to convert anything. You need an answer that's spread across several files. The AI PDF Workspace reads the text of up to 20 PDFs so you can search all of them at once, right in your browser. It can also ask Claude, an AI model made by Anthropic, to compare the documents, list where they disagree or write one summary, with a file name and page number for each point. Search is free and never leaves your browser. AI answers send the text to Anthropic only when you ask and agree, so check each page reference before you rely on it.

Archiving instead of converting

Sometimes the goal isn't to reuse content but to keep the document readable for decades. PDF to PDF/A rewrites a file to the PDF/A-2 archival standard with Ghostscript, embedding fonts and an sRGB color profile so the file doesn't depend on anything outside itself. No converter can guarantee that every input passes strict validation, so check the result with a validator like veraPDF when compliance is mandatory. Repair damaged files first with Repair PDF, give the document a meaningful title with View / Edit Metadata, and make scans searchable before archiving so they stay findable.

Common workflows

Questions about converting PDFs to other formats

Related guides