Edit existing PDF text: change wording, size and colour. Add text boxes, images, shapes, highlights and whiteout, then save. Free, no sign-up, no watermark.
Edit text cursor is always armed, so a single click on any line of existing text turns it into an editable box with a caret in it. Retype the wording, or use the selection popover to change the size and the colour. When you save, the operators that drew the original characters are taken out of the page and your replacement is written in their place — the old text is removed rather than painted over. If a particular line cannot be rewritten that way, the editor says so before you commit and offers to cover and redraw it instead, so you always know which of the two you got.Add text and drag a box where you want it, then type. Size, colour and font come from the popover, the box moves from its border, and Delete removes it. New boxes are written as real text objects — searchable and selectable in the saved file, not a picture of words. A box you leave empty is discarded rather than written into the file.Add image and pick a PNG or JPG to place a logo, a stamp, a photo or a scanned letterhead onto the page. Drag it to move, pull a corner handle to resize, and press Delete to take it out. Images you add sit on top of the page content, so place them over whitespace or over a whiteout block. An image that was already in the document can be deleted but not moved or resized in this version, and it is drawn with a delete-only frame so you can see that at a glance.Shapes opens rectangle, ellipse, line, highlight and whiteout; Draw is a freehand pen for circling a figure, ticking a box or scribbling a correction. Highlight lays a translucent band over text the way a marker pen does, leaving the words legible underneath. All of them are objects on the canvas — select, move, resize, delete, undo.<name>-edited.pdf. The file you opened is left exactly as it was, so you can start over whenever you want. There is no sign-up, no watermark and no charge at the point of saving.Every action is undoable: Undo and Redo sit in the top bar and never move, and Ctrl+Z / Ctrl+Shift+Z (Cmd+Z and Cmd+Shift+Z on a Mac) step back and forward through your recent edits. The history is deep but not unlimited — when it fills, the editor shows an "Undo history is full" notice and folds the oldest edits into the document it is working from. Nothing is lost from the file; those earliest steps simply stop being undoable. Ctrl+F opens Find & replace with per-page hit counts, and a "Show editable areas" toggle outlines every run on the page by what you are allowed to do to it.
Need to change the pages rather than what is printed on them? Rotate PDF, Delete Pages, Reorder PDF and Extract Pages handle the page level. To make the finished file smaller, use Compress PDF; to glue two edited documents together, Merge PDF.
This is a content editor: it changes what is printed on the page. Anything that changes the set or order of pages belongs to the page tools linked further down, and a few jobs are not in this version at all.
| Job | Here? | How it works, or where to go instead |
|---|---|---|
| Change the wording of existing text | Yes | Click the line, retype it; the original characters are removed from the file |
| Change size or colour of existing text | Yes | Select the line, set size and colour in the popover |
| Add a new line or block of text | Yes | Add text, drag, type — written as real, selectable text |
| Add / move / resize / delete an image you place | Yes | Add image; drag to move, corner handles to resize, Delete to remove |
| Move or resize an image already in the document | No | Delete-only in this version, and the frame shows it |
| Rectangle, ellipse, line | Yes | Shapes, with colour, stroke width and opacity |
| Highlight text | Yes | Translucent band drawn over the words, which stay readable |
| Freehand marks | Yes | Draw pen |
| Cover something up, or cover and remove it | Yes | Whiteout, with the choice made at draw time — read the table below first |
| Edit text in a scanned page | No | Nothing to click: a scan is a picture. Run OCR elsewhere first, or whiteout and retype |
| Edit text inside a form field, stamp or comment | No | Detected and named in Document issues, never edited. Flatten PDF turns already-filled values into page content |
| Fill in a form, or sign a document | No | Not in this version |
| Create or edit a hyperlink | No | Links live in the annotation layer, which this version does not touch |
| Re-wrap a paragraph | No | Edits are line-level; a longer replacement grows its own line |
| Rotate, delete, reorder, extract pages | No | Rotate, Delete Pages, Reorder, Extract Pages |
| Watermark, page numbers, N-up layout | No | Watermark PDF, Add Page Numbers, N-up PDF |
| Compliance-grade redaction | No | Whiteout hides, and can remove page text, but it is not a certified redaction workflow |
A PDF does not store paragraphs. It stores drawing instructions: "set this font at this size, move to this coordinate, show these characters". A line of body text is usually one or a few of those show-text operators, and consecutive characters that share a font, a size and a colour on one line are what this editor treats as an editable run. That is the unit you click.
Rewriting a run is a content-stream edit. The editor finds the operators that drew the original characters, removes them, and writes new ones using the same font resource, position and colour unless you changed those in the popover. Two consequences are worth knowing before you start.
pdftotext; it is not there. That is the point of doing it this way rather than drawing a white box and typing on top.Some runs cannot be rewritten losslessly — text converted to vector outlines by the program that made the PDF, text drawn with a Type 3 or a symbol font, text baked into an image, text positioned in a vertical writing mode, or scripts that need a shaping engine this version does not carry. When the editor hits one of those it says so on the spot, names the reason, and offers the fallback: cover the original and redraw your replacement over it. The notice is deliberate. A silent fallback would leave you believing the old words were removed when they were still sitting in the file.
The first thing the editor tries is the font already inside your PDF. If the document embeds Helvetica and every character you typed exists in that embedded font, your edit is drawn in the document's own Helvetica and the line is indistinguishable from its neighbours.
The complication is subsetting. To keep files small, most PDF producers embed only the glyphs the document actually uses — a page that never printed the letter Q has no Q in its embedded font, and a typical subset carries a few dozen characters rather than a few hundred. Type a character the subset does not contain and the original font simply cannot draw it, so a stand-in face is used for that run. In practice that happens on most edits, which is why the substitution is announced rather than hidden: the notice names the face and the reason.
| Original font family | Stand-in used | Why this one |
|---|---|---|
| Helvetica, Arial and other grotesque sans faces | Liberation Sans | Built to the letter widths of Arial and Helvetica, so line lengths hold |
| Times New Roman, Times and other transitional serifs | Liberation Serif | Built to the letter widths of Times New Roman |
| Courier, Courier New, monospaced faces | Liberation Mono | Built to the letter widths of Courier New; column alignment survives |
| Calibri | Carlito | Designed to match Calibri's letter widths |
| Cambria | Caladea | Designed to match Cambria's letter widths |
| Anything else, or a font the file only names without embedding | Liberation Sans, Serif or Mono, whichever is closest | Chosen from the family the original belongs to, so the page keeps its character |
The font picker shows the contract rather than hiding it — entries read "Arial (Liberation Sans)", "Calibri (Carlito)", "Cambria (Caladea)" — and the document's own faces are listed last, under a divider, with an entry greyed out and a reason given whenever it cannot draw everything you typed. The stand-in faces use the same letter widths as the fonts they replace, so a rewritten line occupies the same horizontal space and does not shove the rest of the layout around; where a width does differ, the editor adjusts that run's spacing to the same effect. The letterforms are close but not identical, which is why a substituted run can look very slightly different from the line above it. If that bothers you on a visible heading, restyle the whole run rather than half of it, so the difference reads as a deliberate style rather than a patch.
Some scripts have no stand-in here at all. CJK text is editable only when the document's own font can draw every character you type, and right-to-left and other complex scripts that need a shaping engine are refused with a reason rather than drawn wrongly. Liberation Sans, Liberation Serif, Liberation Mono, Carlito and Caladea are used under the SIL Open Font License, and the licence file for each family ships alongside it.
Everything you did not touch is carried straight through. Images keep their original compression and are not re-encoded, vector art stays vector, tables and rules keep their exact coordinates, hyperlinks and bookmarks survive, page size and rotation are untouched, and a line you did not edit keeps its exact position — editing one line never nudges another. There is no rasterise-the-page step anywhere in this tool, which is the usual reason an "edited" PDF comes back blurry and several times bigger elsewhere.
What does change:
/Artifact marked-content block carrying an /XConvertEdit property. That is how the editor finds and re-selects your own objects if you open the saved file here again; every other viewer and printer treats the block as ordinary page content.New text you add is written as real text objects with an embedded font, so it stays searchable and selectable — it is not a picture of words pasted on the page. Shapes, highlights and free-draw strokes are vector marks and, correctly, are not text.
These get confused constantly, and the confusion leaks confidential documents. Here is the honest version:
| What it does to the file | Can the original still be extracted? | Use it for | |
|---|---|---|---|
| Whiteout, cover only | Draws an opaque block on top | Yes — the text under the block is untouched and any extraction tool will still find it | Tidying up a stale footer or letterhead in a document that is not sensitive |
| Whiteout, cover and remove | Draws the block and takes the covered text out of the page content | No, for the runs it reports as removed — and it names anything it could not remove | Blanking an area, including the invisible text layer a scan carries after OCR |
| Editing the text | Removes the original show-text operators and writes new ones | No — the replaced wording is gone from the page content | Correcting names, dates, prices, typos; taking a sentence out |
| Certified redaction (not this page) | Removes the content and sweeps metadata, attachments, annotations and other hiding places | No, by design and by audit | Legal, medical, FOIA and compliance work |
Taking text off a page removes it from the page it was on. It does not go hunting through document metadata, XMP packets, embedded attachments, annotation objects or earlier incremental-save revisions, and it is not certified for anything. Text inside an annotation or a form field is a special case worth repeating: those appearances are painted after the page content, so a whiteout block written into the page does not even hide them. The editor reports that as "not removed and still visible" rather than letting you assume otherwise. If a document is going out under a redaction obligation, use a tool built and audited for redaction.
Stated plainly, because you should find these out here rather than three edits in:
| Limit | What it means in practice | What to do instead |
|---|---|---|
| No paragraph re-wrap | A longer replacement extends its own line rather than reflowing the block | Edit line by line, or shrink the run to fit when the editor offers |
| A stand-in font on most edits | Embedded subsets are small, so a character you type is often missing from the original face | Nothing — it is announced, and the widths are matched; restyle the whole run if the join shows |
| No complex-script editing | Scripts needing a shaping engine, and vertical writing, are refused with a reason | Use the document's own font where it covers your text, or add a new text box |
| No OCR, forms, signatures or certified redaction | A scan has no text to click; form fields and signature fields do not accept input | Run OCR elsewhere first; Flatten PDF for filled values; a dedicated service for signing |
| Symbol and outlined text are not retypeable | Wingdings-class faces and text converted to vector outlines can be deleted but not rewritten | Whiteout the area and type a replacement over it |
| Page-level work is elsewhere | Adding, deleting, rotating and reordering pages are separate tools | The page tools listed below, used after this one |
Two things this version does better than most of the alternatives are worth saying out loud. The editing work is done on your own machine — see Privacy, below. And Save is not a paywall: there is no export-time upsell, no watermark on the output and no account to create, so the file you can see on screen is the file you can download.
The whole document is parsed and rendered before you can click anything, so file weight costs you time at both ends rather than money. The editor shows its working limits on the upload panel itself, and those figures are lower on a phone or a low-memory device than on a desktop — they come from the tool, not from this page, so what you read there is what is actually enforced. Past the soft line you get a warning and can carry straight on; past the hard ceiling the document will not open here.
For a document past the ceiling, the reliable route is to work in sections: Split PDF into parts, edit the part that needs changing, then Merge PDF the pieces back into one file. A scan-heavy document is often over the line only because of image weight, in which case Compress PDF first will usually bring it back into range without you having to split anything.
This page changes what is printed on a page. The rest of the PDF suite changes the pages themselves, and the two are designed to be used one after the other:
Do the page-level work last, so numbering and stamps reflect the final document. Converting instead of editing? PDF to Word gives you a document you can rework in a word processor, PDF to JPG exports pages as images, and the PDF hub lists every PDF tool on the site.
Processing happens locally in your browser.
There is no account to create and no email to hand over to use this page, and nothing is added to the document you download — no watermark, no "edited with" line, no tracking marker. The saved file contains your document and your edits, and that is all. Your original is not modified either: Save always writes a new file, so the copy you started from is still sitting where it was if you want to begin again. If you close the tab mid-edit, the editor can offer your edits back the next time you open the same document; that record is kept on this device only, is cleared when you save or start over, and can be cleared by hand from the ⋯ menu.
Not the scanned text itself: a scan is a photograph, so every word is part of one image. Clicking opens nothing, and the editor says so.
Depends which tool, and the editor is explicit at the time.
pdftotext over it: the wording is gone.Almost always subsetting: only the glyphs a document uses get embedded, so its "Helvetica" may hold a few dozen characters.
Liberation Sans, Serif and Mono, Carlito and Caladea ship under the SIL Open Font License, each licence file beside its family.
No editable text run sits under your click. The editor names which case you hit, and "Show editable areas" classifies the page — editable, delete-only, read-only. Usual causes:
Workaround: whiteout the area and type over it.
Not as a file. Saving re-serialised the document — the whole PDF is written afresh, so object numbering and byte layout differ even on untouched pages. What is preserved:
Two password kinds behave differently; this version treats both alike on screen.
No. A line you did not edit keeps its exact operators and position; editing one never moves another.
Check at 100% zoom before saving.
Usually not. The editor warns you before navigating away with unsaved work, and after each edit records the changes — not a copy of the file — to offer back when you reopen that document.
⋯ menu.Yes, in any current mobile browser.
No to all four; each has a route.