Harvest all images and text from a brochure
Goal
Take a finished product brochure and pull out everything reusable: the photos as separate files, the copy as plain text, and a record of who made it and when.
The scenario. The marketing agency that made your 2024 brochure has vanished, and the source files with them. A partner wants to reuse the product photos on a website, and a colleague needs the text to start a new brochure. All you have is the PDF.
Starting state
- A PDF. Test: can you select text? If not, it is a scan (see Variations).
- Unprotected (Unlock if needed).
- Respect copyright: the photos and text may belong to the agency or a stock library. Check your rights before reusing them.
Chain
Extract Metadata → Extract Image + Extract Text (or Parser for everything at once) → [OCR Image]
who/when/what the pictures the words text in images
Step by step
1. Extract Metadata: know what you have
- App: Extract Metadata; no options.
- Read the title, author, creation date, application and page count; download
extracted_metadata.csv. - Why: it tells you the source and age of the file. Values may be wrong or default (a username or template name).
2. Extract Image: the pictures
- App: Extract Image; no options.
- You get every embedded bitmap as a PNG (
brochure00000001.png...), as a ZIP when there are many. - Notes: quality is exactly what is stored in the PDF; a small JPEG inside can become a bigger PNG; tiny decorations (bullets, lines) appear too; vector logos and charts are not extracted.
- If the brochure is long, use Splitter first to avoid hundreds of files.
3. Extract Text: the copy
- App: Extract Text.
- Formatting mode: Pure for reading and reuse (default). Flatten if columns come out mixed up (it orders fragments by position on the page). Raw if Pure joins or splits lines wrongly.
- You get one .txt per PDF. Tables and layout are not kept.
4. Alternative: Parser, everything at once
- App: Parser; no options. One ZIP with the text, every picture, embedded attachments,
annotation.xfdf(comments),bookmarks.xmlandformdata.xfdf. - Pick it when you want the whole bundle (an audit of what is hidden in the file). Pick steps 2 and 3 when you want only pictures and words: simpler and lighter.
- Parser does not extract tables.
5. Optional: OCR Image for text inside pictures
- App: OCR Image. Send it a picture where text is part of the image (a screenshot of a price table, a stamp).
- Why: Extract Text reads the PDF's own text only, never the pixels.
Expected result
A folder with the PNGs, a .txt of the copy, and the metadata CSV (or one Parser ZIP).
Check it worked
- Count the product photos in the PDF and compare with the number of sensible PNGs after you discard icons.
- Open the .txt: are headings and paragraphs in sensible order? If not, run again with Flatten.
- Compare the largest PNG to the picture on the page: resolution cannot be better than what was stored.
Variations
- Brochure is a scan: you get one full-page image per page and an empty text. Run Searchable, then Extract Text; send the page images to OCR Image if you only need text.
- Selected pages only: Splitter, range into a single file.
- Rebuild a clean PDF from chosen images: Merger with the images.
- A photo sheet: the collage recipe.
What can go wrong
- No images found. The PDF contains only text and vector graphics.
- Hundreds of files. The brochure is long, or each page is one image (a scan).
- Pictures look low quality. They are stored that way.
- Empty .txt. A scan: make it searchable first.
- Garbled characters. Odd fonts or damaged encodings.
- Results disappear. They are kept about two days; download them.
Why this order
Metadata first because it costs nothing and sets expectations. Images and text are independent, so their order does not matter, but both come before OCR Image, which needs the pictures you just extracted.