Fix what is printed
Click a line of the document’s own text and retype it. A misspelt name, a wrong date, a stale address.
PDF editor
Change the page, not the document
Retype a line that is already printed, add text and shapes, cover what should not be there, and download it. Everything is written into the page the file already has, so the result is still the document you started with.
Edits the page’s own text, in its own font
Runs in your browser, no install
Says what it cannot do instead of faking it
What it is for
The jobs that have nothing to do with AI
A wrong date on a certificate. An account number that should not be in the copy you are emailing. A signature block that needs to go on page three. None of that needs a form to be understood, and none of it should cost a credit.
Add what is missing
Text boxes, rectangles, lines, ticks and signature images, placed where you want them and moved afterwards.
Remove what should not be there
Delete lines outright, or cover an area in the page’s own colour. The editor is explicit about which of the two you are doing.
Reorganise
Add and remove pages, and check every change against the page as it will actually print.
The difference
Editing a PDF, or replacing it
Most free editors import your file, convert every line into their own text objects and generate a new document. It feels the same until somebody compares the result with the original.
Converting the file
- Text re-typed into the tool’s own objects and re-exported
- Fonts substituted for whatever the tool has
- Vector artwork flattened to an image
- Every change appears to succeed
- Output that resembles the original
Editing the file
- The page’s existing operators rewritten in place
- The document’s own embedded font, or an honest refusal
- Artwork, lines and stamps carried through untouched
- Changes it cannot make safely are declined and explained
- The original, with your change in it
Any editor can put your text on a page. The useful question is whether it tells you when the result will not match the document it came from, because the office receiving it certainly will.
Why refusals are the feature
How an edit works
Four steps, none of which regenerate the page
The page is read into its operators
Every text-showing instruction on the page is indexed with its position, font, size and colour, including the ones nested inside reusable page objects.
A line you see may be one instruction or forty; producers that emit one glyph at a time are common, and are joined back into a line you can click.
You retype the line
Your text is encoded with the codes the document’s own font already uses, so the characters map to the glyphs actually embedded in the file.
This is where a refusal comes from: a character the file has no glyph for cannot be drawn in that font, and pretending otherwise is what produces the wrong-looking result.
The bytes are spliced
Only the affected instruction is replaced. Position, spacing and clipping are preserved, and every other byte of the page is left exactly as it was.
No re-encode, no regeneration, no reflow of anything you did not touch.
You review the real page
The result is rendered and shown to you before you download it, so what you check is the printed page rather than a preview of the intention.
Deleting
Hidden and removed are not the same thing
This matters more than it sounds. A black rectangle over an account number in a PDF is not redaction: the number is still in the file and any reader can select it.
Delete removes it
The glyphs come out of the page’s content stream. Afterwards the words are not in the document, and selecting or copying that area finds nothing. This is what you want before sending a copy of something.
Cover hides it
A rectangle in the page’s own sampled colour is placed over the area. Used where the text cannot be rewritten safely, or where it is part of a picture. The text stays in the file, and the editor says so rather than letting you assume.
Where it stops
What it will not do, and why
Written down rather than discovered, because a limit you know about is a decision and a limit you meet halfway through a job is a waste of an afternoon.
Characters the file has no glyph for
PDFs embed a subset of a font, not the whole thing. A letter the document never prints cannot be drawn in that font. You are offered a cover and a text box instead.
Text that is part of an image
A scanned or stamped-in heading has no letters to edit. Right-click it and the editor says so, and offers to cover it and type over.
Shared page objects
Where one instruction is drawn in more than one place, changing it would change both. That is refused rather than done quietly.
Active content and dynamic forms
PDFs carrying JavaScript, launch actions or embedded files are refused on upload, and dynamic XFA forms are declined because their real content is not on the page at all.
Where it runs
On your machine, in your browser
0installs
0credits to edit and download
1file changed: yours
Rendered locally
The page is drawn by a PDF engine compiled to WebAssembly, in a worker thread, on your own machine. Your document is not handed to a third-party viewer to be displayed.
Stored encrypted, deleted when you delete it
Files you keep are encrypted at rest for your account. Nothing is used as training data, and nothing is submitted anywhere on your behalf.
If the difficulty is not changing the page but understanding what the form is asking for, that is theAI document editor.
Start here
Open a PDF and change one line
No AI, no credits, no install. See what it does with your document.
- Reading a form and reviewing the result cost nothing
- Five free credits, no card
- Your file is deleted after 30 days unless you keep it