The PDF Tools Everyone Should Know

PDFs are the format everyone shares and nobody enjoys editing. Almost everything you will ever need to do to one comes down to six operations. Learn to do them cleanly and the format stops fighting back.

Two things decide most of the outcome, and neither is the tool you pick. The first is where the file gets processed. The second is what order you do things in, because one of these six operations destroys the results of another if you run them the wrong way round.

Where the file actually goes

Splitting, merging, and converting to images or text are all arithmetic on the file's own structure, and they run inside the browser tab with the document never leaving your machine. Compression is different: it needs a full PDF interpreter, which means a server and an upload.

That is not a philosophical distinction. It changes who has a copy. On this site, PDF Merger , PDF Splitter , PDF to Image and PDF to Text run locally; PDF Compressor uploads. Why Your PDF Tool Should Run in the Browser covers what an upload exposes and how to check any tool for yourself in about a minute.

Shrink a file that is too big to send

The classic blocker: the PDF is 28 MB and the upload caps at 10.

Before reaching for a slider, work out why it is heavy, because that decides whether compression can help at all. Open it and scroll. Photos and scanned pages mean the images are the cost. Mostly text and still huge means embedded fonts, and compression will barely touch it.

That is not a hedge, it is how the presets work. The four settings every Ghostscript-based compressor exposes are image resolution settings: screen downsamples colour and greyscale images to 72 dpi, ebook to 150, printer to 300, and prepress also to 300.¹ If the file has no images, there is nothing for any of them to take away.

So the honest advice is short. Ebook, 150 dpi, for anything read on a screen. Printer, 300 dpi, for anything that gets printed. Screen at 72 dpi is a last resort when a hard size cap leaves no choice.

Two things worth knowing that the labels do not tell you:

Prepress is not the quality setting. It is 300 dpi, the same as printer, and Ghostscript's own documentation says plainly that prepress "does not indicate the highest quality conversion" and that using any preset alters the input.¹ Choosing it because it sounds most careful usually just produces a larger file.

The size drop is geometry, not magic. Halving the resolution quarters the pixel count. Going from 600 dpi to 150 is a quarter of the linear resolution and therefore a sixteenth of the pixels, so an image-heavy scan can genuinely shed most of its bulk. How much of the file that removes depends entirely on what fraction of it was images to begin with, which is why the same setting can take one document from 30 MB to 2 MB and another from 30 MB to 29 MB.

Whichever you pick, check the result rather than trusting the number. Open it, zoom to 200 percent on a page with small text, and look. Crisp means you had room to spare. Mushy means step back one preset.

Merge and split without scrambling pages

Merging is assembly: cover letter, resume, portfolio into one file. The thing that trips people is order, because tools concatenate in list order, not filename order or the order you dropped them in.

Look for reordering controls before you resort to renaming. Ours puts up and down arrows on each row and merges top to bottom exactly as shown. Renaming to 1-cover.pdf, 2-resume.pdf is the fallback for tools that sort by name and give you nothing else.

Splitting is the reverse, and there are two jobs: pull one page out of a contract, or cut a long report into sections. The feature to look for is page-range syntax, so you can say 1-5, 12, 20-40 in one pass rather than deleting pages one at a time. PDF Splitter takes that syntax and shows the resulting page count as you type, which catches an off-by-one before you export rather than after.

Convert into and out of PDF the right way

"Convert" means two different jobs with different difficulty.

Going into PDF, from a document or an image or a slide deck, is about freezing layout so it renders the same everywhere. This is usually painless.

Going out is where quality decisions live:

• To image. PNG for text and sharp lines, because it is lossless and edges stay clean. JPG only for photo-heavy pages where file size beats perfect edges. Watch the scale control, which multiplies the page's natural size: 2x is a reasonable default for reading on a screen, higher if you plan to crop into a detail.
• To Word or text. This works only if the PDF holds real, selectable text. Try to select a sentence. If your cursor highlights it, conversion keeps the text editable. If nothing selects, you have a scan, and no converter can help until you deal with that.

Make a scan searchable with OCR

OCR reads the pixels of a scanned page and writes an invisible text layer underneath, which is what makes a scan searchable, quotable and convertible.

The quickest test for whether you need it: drop the file into PDF to Text . It reads the existing text layer and nothing else, so a scan comes back empty. That is a definitive answer in one step rather than a guess from squinting at the page.

If you do need OCR, input quality decides the result more than the engine does, and the guidance is more specific than "scan at 300 dpi":

• 300 dpi is the floor, and it is the minimum Tesseract itself recommends.² But dpi is a proxy. What actually governs accuracy is how tall the letters are in pixels, so a 300 dpi scan of tiny print can read worse than a 200 dpi scan of large print.
• Straighten the page first. Skew degrades line segmentation before recognition even begins,² which is why a slightly rotated phone photo produces such disproportionately bad output. Deskewing buys more accuracy than any amount of extra resolution.
• Leave a border. Text running to the very edge, and equally a small amount of text marooned on a huge background, both cause problems.²

Then confirm the result the same way you tested for it: search the document for a word you know is on the page.

Sign last, and know which kind you need

Two different things share the word "signature".

A visual signature drops a picture of your handwriting onto the page. It proves nothing and stops nothing; it is fine for an internal approval where the point is showing intent, not proving it. Sign a white sheet, photograph it, crop to a transparent PNG, and keep that file for reuse.

A digital signature binds a certificate to the document cryptographically and flags any change made afterwards. That is what you want for anything legal or financial, and it needs a certificate you or your organisation holds.

Which brings back the ordering problem. A PDF digital signature covers a declared byte range of the file as it stood at signing, and later edits survive only when they are appended to the file as an incremental update.³ A compressor does not append anything: it interprets the document and writes a new one from scratch. Every byte offset moves, and the signature over the old layout is dead.

So: compress, merge, split, convert, OCR, and only then sign. It is the one hard ordering constraint among the six, and it is the one people discover after the fact.

Master these and the format stops being a chore.