76 guides · three families · written by the team
Every other guide stops at “open your PDF editor”. These start with the actual menu path on the device you are holding, then explain what the operation does to the file — including the parts that go wrong.
How these are written
Each guide is written and reviewed by the Ream team, checked against the current iOS, Android, Windows 11 and macOS behaviour, and dated. Where a tool cannot do something honestly, the guide says so.
30 guides
Where the button actually is on iPhone, Mac, Windows, Android and Drive.
The iPhone opens, signs and emails PDFs, but Apple never shipped a merge command. Here is what the Files app really offers and how to combine documents on the phone itself.
Preview has had a merge feature for years and almost nobody knows where it is. Drag pages between two thumbnail sidebars, then save — and mind what the drag does to links.
Windows can open a PDF and print one, but it cannot join two. This walks through the print-dialog myth and the browser route that actually produces a single document.
Android has no system merge command and its file managers are browsers, not editors. The fastest route is a browser tab that reads both files from the storage picker.
Google Drive is storage, not an editor. Merging from Drive means using a tool that reads the mounted file directly — or accepting the Docs detour and its damage.
Apple’s PDF tools stop at annotation, so splitting means a browser tab and a clear idea of which pages you actually need. Page labels make that harder than it sounds.
Preview hides a compressor in its Export sheet. It works, up to a point — usually the point where your 300 DPI scan becomes a 72 DPI smudge. Here is how to check.
Markup holds a saved signature and stamps it onto any page. It takes about twenty seconds, needs no app, and is an electronic signature rather than a cryptographic one.
Every Windows 11 route called combine turns out to be a print job or a rename. This maps the shell honestly, then combines the files in a browser tab with no install.
Android has no PDF compressor, so a phone is the worst place to fix a 40 MB scan — unless you use a browser tab that knows the difference between an image and a word.
Preview cannot export DOCX and Pages will not convert a PDF’s text. The honest options are a browser converter, Google Docs, or LibreOffice — each loses something different.
There is no Windows setting for PDF passwords, and a password-protected zip is a different thing. This covers real AES-256 protection, permissions, and key handling.
macOS has no Split command. Preview subtracts pages, Shortcuts can batch the job, and a browser tab is the only route that copies pages without re-rendering them.
Edge cannot shrink a PDF and neither can the shell. The routes that work are Office’s minimum-size export, Ghostscript from a terminal, and a browser compressor with a real DPI choice.
The iPhone has no PDF compressor: Compress makes an archive, and Mail has no attachment-size option for documents. Safari with a local compressor is the route that keeps the file on the phone.
Android has no signing tool in the system, so the choice is a vendor app or a browser tab. Both produce an electronic signature; neither produces a certificate-backed one.
An app is the right call when you merge weekly or work offline. It is the wrong call when the app uploads, watermarks or asks for access to every file on the phone.
iOS will produce a PDF from photographs in about six taps, using the print preview nobody knows is interactive. It paginates one photo per page and strips most metadata on the way.
macOS has three built-in image-to-PDF routes and none of them asks about page size. Knowing which one produces a printable page rather than a 56-inch sheet is most of the work.
The best free route on Android is Drive’s built-in scanner, which crops and straightens as it captures. The fastest is a browser tab, and the worst is a print-to-PDF of a photo.
macOS has no page-numbering command: Preview stamps one page at a time and the print sheet has no footer field. A browser tool or a short script is the honest route.
Windows has no watermark command, and Word’s version only works on documents you still own. A browser stamp or a PowerShell command are the two routes that work on any PDF.
iOS Markup rotates one page at a time and saves the change on Done. For a batch, a browser tool writes the rotation attribute into every page in one pass.
Windows has one built-in way to get pages out of a PDF: print a range. It works and it degrades. A browser extract tool copies the pages instead, which is a different file.
iOS has no print drivers. If the printer does not advertise AirPrint over Bonjour, the document has to reach it through a vendor app, Wi-Fi Direct, or a Mac acting as a bridge.
The attachment limit belongs to your mail account, not the phone, and base64 adds a third to the file. Mail Drop, a Files link or a split are the three honest escapes.
A phone can read text off a page in seconds, but reading it is not the same as writing a searchable layer into the file. The difference decides what you can do afterwards.
Preview’s crop tool works per page and sets a page box rather than deleting content. For all pages, or for numeric margins, a browser tool is the faster route.
Files cannot make a form field and Docs cannot export one. A browser form tool writes AcroForm fields on the device, which is what every reader understands.
Firefox and Safari both run a local merge without an extension, but neither offers a save dialog, and Safari may unzip the result for you. Knowing the differences avoids a lost file.
25 guides
The document problems people search for at 11pm, solved properly.
Nearly every oversized PDF is a scan of paper pretending to be a document. The others are fonts, duplicate saves, attachments and images at print resolution.
Search, copy and screen readers all depend on a text layer that a scan does not have. OCR adds one without touching the image underneath.
If you can open the file, you can usually save an unencrypted copy. If you cannot, no online tool is going to help — and most that claim to are lying.
A PDF page carries a rotation value that viewers apply on the fly. Set it properly and every reader agrees; rotate the pixels instead and you lose quality.
Drawing a rectangle hides text from eyes, not from software. Proper redaction removes the characters, the layer they sit in, and every copy hiding in metadata.
The upload is invisible, which is why nobody thinks about it. Here is what to look for in the network panel and why aeroplane mode is the only test that cannot be faked.
Mail servers measure the encoded message, not the file, so the ceiling arrives sooner than you expect. Compress, split, or stop emailing it altogether.
Fillable means interactive fields that a reader can tab through. Adding them is straightforward; making them appear correctly in every reader is the part that takes care.
A table in a PDF is not a table; it is text placed at coordinates. Getting it into rows and columns means reconstructing structure that the file never stored.
Most broken PDFs are missing bytes at the end, which a repair can often work around. If the page streams themselves are gone, no tool will invent them.
Most courts want a text-searchable PDF within a size limit; some archives and local rules require PDF/A specifically. The conversion is easy; the validation is the part that matters.
Choose the scheme before the font: stamped numbers, page labels and printed folios are three different things, and a document usually needs two of them to agree.
Documents travel through WhatsApp untouched; photographs are compressed to about 1,600 pixels. Most lost quality happens before the file is shared, not during the send.
A contract changes in three ways: words, layout and metadata. One comparison cannot see all three, so a proper check runs two diffs and reads the parts a tool cannot judge.
Portals measure megabytes, not legibility. Find out what is in the file first, then reduce the right layer: image resolution, then pages, and text never.
A bookmark points at a page; a splitter counts pages. Translating one into the other is a five-minute job if you note the physical page of every top-level entry first.
Language packs are sets of shapes plus a dictionary. Pick the script, not the country, and check the numbers and names that the dictionary cannot repair.
Two metadata systems, four places people forget, and one reliable local removal: rebuild the file. The cost is the text layer, so decide before you send.
Two documents, one file, and a numbering decision that cannot be reversed cheaply. Continuous, prefixed or original folios: each is defensible, and mixing them is not.
A cryptographic signature covers a byte range, not the whole file. Read which revision it protects, then compare copies when the signature has nothing to say.
A converter gives you rows and columns. Making the sheet editable means fixing types, stitching pages and rebuilding the calculations the PDF never stored.
A handout is not the deck printed smaller. Choose slides, notes or outline first, then match the paper, the resolution and the ink to what the room needs.
A scanned blank page is never blank: noise, bleed-through and shadows all count as ink. Judge pages visually, remove them, then prove nothing was lost.
Accessibility comes from tagging at authoring time. OCR adds words, not structure, and PDF/A-2b is an archival profile, not an accessibility one.
Files fail because they depend on something outside themselves, or because nobody checked the copies. Convert, validate, record a checksum, store three, verify on a schedule.
21 guides
The format underneath the button, explained without hand-waving.
They are not two versions of the same document. One describes marks on a page, the other describes content that flows into whatever page it finds.
Recognition is a chain of guesses: find the text, split it into lines, split lines into characters, classify each one, then repair the result with a language model.
An electronic signature is a legal concept. A digital signature is a mathematical one. A drawn signature can be the first, but it is never the second.
Embedding is optional, subsetting is common, and the standard 14 fonts are assumed to exist everywhere. Break any of those assumptions and text arrives as boxes.
There are two ways to make a PDF smaller: remove redundant structure, or replace the page with a picture. One keeps text; the other does not.
The distinction decides whether search, copy, accessibility and redaction will work at all — and why two files with the same page count can differ by a hundred megabytes.
A password derives a key, the key encrypts streams and strings, and a permissions flag list asks readers to behave. Only the first part is a lock.
The “2” is the generation, the “b” is the visual-fidelity level. It is the pragmatic archival target for modern documents, not a certificate of correctness.
Outlines live once per document and address pages by reference. Copying pages without copying and remapping the outline is why a 300-page manual arrives with no navigation.
Every PDF page carries up to five nested boxes. Confusing them is why a crop does not stick, why a printer asks for bleed, and why a page looks smaller than it is.
A browser has no PDF engine built into HTML. Each one ships or bundles a renderer that paints pages as pixels and fakes text selection with a positioned layer.
There are two metadata stores and two kinds of compression, so the answer is “it depends which one”. Neither is a reliable way to scrub a document.
Two operations hide behind one button: re-encoding images, or re-rendering the page. Knowing which one runs tells you whether the text survives.
Blur is a resolution problem decided at scan time. Effective DPI, viewer zoning and JPEG artefacts explain it, and upscaling cannot undo it.
Merging layers makes optional content permanent. Flattening can also mean compositing transparency, baking forms or burning in annotations. The consequences differ.
Nothing is broken. The printer fits the page into what it can print, or the paper is a different size, or the page box is larger than the visible page.
Values live in field dictionaries, appearances are drawn from them, and old revisions can keep earlier answers. Flattening removes the field, not the words.
Integrity is the strong claim, identity is a certificate authority’s claim, and time needs a timestamp. A valid signature is not the same as a trustworthy document.
Pagination is a result in Word and a stored fact in PDF. Matching the page setup removes most of the drift; the rest comes from guessed structure.
Text pages are kilobytes of instructions; scanned pages are millions of pixels. Dividing size by page count tells you which one you have.
A PDF freezes a page. That is exactly right for a contract and exactly wrong for a budget, a web page or a document someone must edit.
Reference
AcroForm, XFA, JBIG2, trim box, subsetting, PAdES. If a guide uses a term and does not explain it, the glossary does.
Read the glossaryStraight answers
Every claim here is testable: the UI paths were checked on current versions, and the tool limitations are the ones the code actually has. If something has moved, tell us at hello@ream.tools and it gets fixed with a new review date.
why local processing matters