Translate PDF

You get text, not a translated PDF

The most important distinction on this page, because it decides whether this tool or a different one is what you want.

What you get here: the words of your document, translated, in a box you can copy or download as a text file. Paragraphs are preserved in order. The layout is not — no columns, no headings as headings, no tables, no images, no page design.

What you may have wanted: the same PDF, looking the same, with the words in another language. That is document translation rather than text translation, and it is a substantially harder job — the text has to be replaced in place, re-flowed to fit boxes that are now a different length, with fonts that support the new script.

If that is what you need, use a tool built for it. DeepL's document translation and onlinedoctranslator both return a PDF with the layout preserved, and DeepL in particular is very good at it. Neither is this tool, and saying so is more useful than pretending otherwise.

Text is the right output when you need to read a document rather than reproduce it · you are quoting from it · the content is going into an email, a report or a summary · you want to search it · or you simply need to know what it says before deciding whether a proper translation is worth paying for. That last case is very common and this tool is well suited to it.

Long documents are handled properly

Worth explaining, because this is where most free translators fail quietly.

There is a limit to how much text a language model can translate in one request. The usual consequence is that a long document comes back cut off — and that failure is close to invisible, because if you could read the target language well enough to notice the ending was missing, you would not be translating into it.

This tool avoids that in three ways.

It splits the document into sections and translates each one separately, so the limit applies per section rather than to the whole document. In total it handles around 60,000 characters — roughly twenty to thirty pages of ordinary text.

It splits on paragraph breaks, so a section never ends mid-sentence. That matters for quality as well as tidiness: a model translating a complete paragraph does a better job than one translating half of one.

And if your document exceeds even that, it tells you — a note is added to the end of the output saying how much was translated. So you always know whether you have the whole thing, which is the part that matters.

Two practical consequences. A long document makes several requests rather than one, and the allowance is thirty requests an hour — so a single long PDF can use a fair share of it. And it takes proportionally longer, because the sections are translated in turn.

For anything over about thirty pages, split the PDF first and do it in parts.

A scanned PDF will produce nothing

The commonest reason this fails, and it takes a second to check.

A PDF either contains real text or a picture of text. A scan or a phone photograph of a page is the second kind — the page is one image and there is no text in the file at all.

This reads the text layer, so a scan gives it nothing to translate.

The test: open the PDF and try to select a sentence with your cursor. If words highlight individually, you are fine. If the whole page highlights as one block, or nothing does, it is a scan and you will need optical character recognition first.

There is a second, subtler limitation from the same source. A two-column document interleaves — text is read in the order it is stored rather than the order you read it, so a line from the left column is followed by the line beside it from the right. Academic papers, newspapers and two-column reports come out scrambled, and translating scrambled text produces scrambled translation. If your PDF is in columns, extract it one column at a time or use a tool that understands layout.

Do not translate anything confidential

Being direct about this, because most tools on this site behave differently and the difference matters.

The text extraction happens in your browser — your PDF file itself is never uploaded. But the extracted text is sent to a language model to be translated, because that is where the translation happens.

So this is not a private tool in the way the file converters here are. Do not put a contract, a medical letter, personal data, unpublished work, legal papers or anything under a confidentiality obligation through it. For those, you need a translation service with an agreement in place, or a human translator.

For a public document, a news article, a manual, a paper, a form you have been sent or a letter you need to understand, it is entirely appropriate.

Getting a usable result

Choose the target language deliberately. Chinese is offered as Simplified and Traditional, and they are not interchangeable — Simplified for mainland China and Singapore, Traditional for Taiwan, Hong Kong and Macau.

Check the ending. Compare the last sentence of your PDF with the last sentence of the translation. Do the numbers and names line up? That takes seconds and confirms you have the whole document even in a language you cannot read.

Spot-check the numbers. Figures, dates, currencies and names should pass through unchanged. They usually do, and when they do not it matters more than any wording problem — so scan the output for the numbers you care about.

Do not rely on it for a decision that matters. Machine translation fails in ways that are invisible from your side: a word taken in the wrong sense, a negation dropped, a technical term rendered as the everyday word. If the document is legal, medical, financial or contractual, use the translation to understand roughly what you have, then get a professional one before acting on it.

Related tools

Extract Text from PDF gives you the PDF's text without translating it, in plain text, Markdown or JSON. AI Text Translator translates text you paste in, with a formality control for languages where politeness is grammar. PDF to TXT Converter is the simplest way to get the words out for reading or pasting.