DeepL Document Translation Limits for Long PDFs and Books

DeepL limits every document twice: by file size and by characters. As checked on October 11, 2026, DeepL's file formats article lists these PDF limits for current plans: 5 MB and 100,000 characters with a free DeepL account, 30 MB and 1 million characters on Pro Individual and Team, and 100 MB and 1 million characters on Pro Business. Word files (.docx/.doc) have the same limits on each plan. The limits apply per file.

For a long book, file size is often not the first limit you hit. Three other numbers matter more: the character cap per file, the number of file translations your plan includes each month, and, on the API, a 50,000-character billing minimum per PDF or DOCX. The table and worked example below show how these combine for a book-length file.

This page covers DeepL only. For Google's document limits, see the Google Translate PDF guide. For a comparison of PDF translation tools, see best PDF translation tools.

DeepL limits by plan for PDF and Word files

All figures come from DeepL's own pages, checked October 11, 2026. Per-file limits come from the support article on file formats and the developer page on API usage limits. Monthly file counts come from the comparison table on DeepL's Pro pricing page. Monthly character allowances for the API come from DeepL API plans.

Notes on reading the table:

• No file count on the API does not mean free files. DeepL's file translation billing article says API plans have no limit on the number of files, but each translated file is billed.
• API Free and API Pro can no longer be purchased. DeepL's API plans article says this. New sign-ups see Developer and Growth. The developer page still labels its columns "free plans" and "paid plans". For PDFs, those columns use the same numbers as the table above.
• Older Pro plans have their own limits. The file formats article lists Starter at 10 MB, Advanced at 20 MB, and Ultimate at 100 MB, all with 1 million characters. Check which plan your account is actually on.
• Choosing more than one target language uses up more files. DeepL's file translation instructions say each target language deducts one file translation from your monthly count. Unused file translations do not carry over to the next month.

We left out prices on purpose. DeepL's pricing page shows a regional price, so any number we printed would only apply to one market. Check your own region on the pricing page.

Counting characters in your book before you upload

DeepL counts characters in the source text. One character is one Unicode code point, and spaces, tabs, and line breaks count too. This is set out in DeepL's usage count and billing article. So "é", "あ", and "深" each count as one character, and so does every space between words.

To estimate a book's count:

1. Get the text into an editor that reports characters including spaces. In Microsoft Word this is the word count dialog. For a text PDF, copy a few representative chapters into the editor.
2. Measure a few typical pages and work out an average per page. Do not use a single dense page.
3. Multiply the average by the number of pages you plan to translate.
4. Count everything that stays in the file. DeepL's document translation guidance says all content in an uploaded file counts toward billed characters, including content that was not actually translated. If you do not need the index, bibliography, or appendices, remove them from the copy you upload. They still count if they are in the file.

Your editor's count and DeepL's count may not match exactly. Leaving some room under each limit protects you from a small difference.

Worked example: one 400-page book on each plan

Assumptions (yours will differ): an English-language, text-based PDF of 400 pages at about 2,000 characters per page, including spaces. We chose that density only to make the arithmetic easy to follow. Measure your own pages before relying on these results.

Total: 400 pages × 2,000 characters = 800,000 characters.

Free DeepL account: 100,000 characters per file, 1 file per month

• 800,000 ÷ 100,000 = 8. Eight files would each sit exactly at the limit, which leaves no room for a counting difference.
• If you aim for about 90,000 characters per part: 800,000 ÷ 90,000 = 8.9, so you need 9 parts.
• With 1 file translation per month, that is 9 months, and the translated files are not editable.

In practice, the free account is for sampling a chapter, not for translating a book.

Pro Individual: 1 million characters, 30 MB, 3 files per month

• 800,000 is under 1 million, so the book fits in one file by character count.
• File size is a separate check. Suppose your PDF is image-heavy and weighs 75 MB (this is a hypothetical size). Then 75 ÷ 30 = 2.5, so you need at least 3 parts. That uses all 3 of the month's file translations, and you have none left for a re-run.
• Team has the same 30 MB limit but more files per month. Business allows 100 MB per file, so this 75 MB PDF would fit in one upload.

API Developer: 500,000 characters per file, 1,000,000 characters in total

• 800,000 ÷ 500,000 = 1.6, so you need 2 parts. Two volumes of about 400,000 characters each are billed at their actual count, 800,000 characters. That leaves 200,000 characters of the Developer allowance, and the allowance never resets.
• Now suppose you split the same book into 24 chapter files instead. Each chapter has about 800,000 ÷ 24 ≈ 33,333 characters. Every PDF or DOCX is billed at least 50,000 characters, so the total becomes 24 × 50,000 = 1,200,000 billed characters. That is 50% more than the text contains, and it is more than the whole Developer allowance.

API Growth or legacy API Pro: 1 million characters, 100 MB

• The book fits in one file if it stays under 100 MB. It is billed 800,000 characters.
• DeepL says a Growth plan on monthly billing includes 1 million characters. This book uses most of a month's included characters.

Legacy API Free: 500,000 characters per file and per month

• You need 2 parts. Together they need 800,000 characters, which is more than the 500,000 allowed each month. So the translation is spread over 2 billing months, for example one 400,000-character volume each month.

When splitting is unavoidable

If the book is 700 pages at the same density, it has 1,400,000 characters. That is above the 1 million-character limit on every plan, so you must split it. The point of the example is the order of the checks: characters first, then megabytes, then the monthly file count, then the billing minimum.

How to split a book that is over the limit

The aim is to make as few parts as possible, each under both limits with room to spare, and then confirm that nothing was lost where the parts meet.

1. Plan volumes, not chapters. Group chapters into the fewest parts that stay under about 90% of your plan's character limit and below its size limit. On the API, keep each part above 50,000 characters, because anything smaller is billed as 50,000.
2. Cut at chapter boundaries. Never split in the middle of a paragraph, table, or footnote block. Notes that sit at the end of the book need a decision. Either upload them with the part they belong to, or translate them as their own part and record that you did.
3. Keep a split manifest. For each part, record the page range, chapter titles, character count, and file size. You will use it to check the translated parts and put them back together.
4. Use the same settings for every part. Use the same source and target language, the same formal or informal tone, and the same glossary. DeepL supports glossaries in file translation. Without one, a name or term can be translated differently in each volume. Our terminology consistency guide explains how to build that list before you start.
5. Check the ends of every part, not just the start. DeepL's file translation troubleshooting page warns that a file over the supported limits might not be fully translated. For each part, open the last page of the output and compare it with the source.
6. Check where the parts meet. Look for chapter headings that were dropped, footnote numbering that restarts, running headers, and cross-references that now point to another volume. A table of contents with page numbers will no longer match after the split, so rebuild it after you join the parts.
7. Run a full QA pass on the joined book. Use the PDF translation QA checklist to sample meaning, tables, captions, and missing content across the whole book.

Things that matter as much as the limits

• Scanned PDFs. DeepL says PDF conversion relies on optical character recognition and may have a higher error rate. It recommends scanning at 300 DPI. A scan that fits the size limit can still produce poor text. If you have a scanned book, read how to translate a scanned PDF first.
• Use the original file if you have it. If a PDF was exported from Word or PowerPoint, DeepL suggests translating the source .docx or .pptx instead. DeepL also says PDFs exported from CAD files are not supported, and password-protected files can't be translated.
• Getting an editable file. DeepL's PDF translation article lists Pro Advanced, Ultimate, Team, and Business as the plans that can download a translated PDF as .docx. Pro Individual is not on that list. On the API, output_format=docx returns a Word file from a PDF. An editable file makes repairs after translation much easier.
• API downloads expire. DeepL's developer guidance says translated documents are stored only briefly and must be downloaded promptly.

When DeepL is the wrong tool for a whole book

If the book fits one or two files on your plan, DeepL's own workflow may be all you need. Run the QA steps above and stop there. Look elsewhere when splitting becomes the main job: many parts, monthly file counts that stretch the work over several months, or API minimums that inflate the cost.

BookTranslator's PDF translator takes a book as a single upload. The upload form accepts files up to 300 MB. It prices by word count and shows the price before checkout. Guest tasks are capped at 300,000 words. Credit-based and Pro accounts have no word limit. A long book is still a long translation, though, so review it. For scanned books, OCR mode rebuilds a clean translated PDF from the recognized content. It does not match the original page layout, fonts, or font sizes. If keeping the layout is the priority, start with how to translate a PDF while keeping the format.