PDF Document Translator
Translates PDFs up to 20 pages (up to 25 MB): pdf.js extracts text in the browser, Gemini AI translates in batches, PyMuPDF rebuilds the PDF (redact + insert into the same zones). Word (DOCX) via pdf2docx from the original with translation substitution; also DOC, HTML, TXT. History stays in the browser only.
Drag a PDF here or click to select
Up to 20 pages · up to 25 MB · text layer · final assembly via PyMuPDF
Progress
Ready to work
Processing history
Stored only in your browser (no sign-in). Up to 12 recent files.
No saved results in this browser yet.
Instructions
- Upload a text-based PDF (up to 20 pages, up to 25 MB). If it's a scan without OCR, recognize it separately first.
- Choose the target language and format: PDF, DOCX, DOC, HTML, or TXT.
- Click "Translate": text is extracted in the browser (pdf.js), and structured text is sent to Gemini in batches.
- PDF: final assembly on the server via PyMuPDF, images/graphics remain, old text is redacted, translation goes into the same blocks.
- DOCX: layout via pdf2docx from the original PDF + substitution of translated lines; DOC/HTML/TXT, separate exports (DOC, flow/fallback if no converter).
- The result appears in the preview (for PDF/HTML) and in "Processing History" without reloading the page.
- Privacy: only text is sent to the AI during translation. The original PDF is sent once for the final assembly of the desired format and is immediately deleted from temporary files.
Повний опис
Professional intelligent PDF translation service with full layout preservation
Translate complex documents, technical manuals, and reports using Gemini AI without losing tables, diagrams, or original formatting.
1. Deep dive into the problem: why does regular PDF translation "break" documents?
The PDF format stores text at strict spatial coordinates. When you copy content into a regular translator (Google Translate, DeepL, etc.), two main problems occur:
- Loss of structure: The translator outputs a solid block of text, completely destroying tables, columns, figure captions, and original formatting.
- Layout collapse when pasting back: If you try to paste translated text into the original file manually, word lengths change (for example, Ukrainian phrases are usually 15–30% longer than English ones), causing text to overflow blocks, overlap images, or "break" table cells.
2. Innovative hybrid processing pipeline
Our service combines the speed of browser technologies, the flexibility of artificial intelligence, and the precision of server-side graphics engines:
- Content extraction in the browser (PDF.js): The document is analyzed directly in your browser. Text is split into logical batches for sending to the AI.
- Intelligent translation (Gemini AI): Gemini AI translates text with context awareness, technical terminology, and preservation of special structure tags.
- Final server-side assembly: When you click "Translate", you get a file that is rebuilt on the server using professional algorithms.
3. Precise format reproduction: PDF, DOCX, DOC, HTML, and TXT
The service offers several output scenarios depending on your follow-up tasks:
- Translated PDF (PyMuPDF Redaction): A specialized PyMuPDF engine is used. The original text is carefully removed ("redacted") from the exact page coordinates, and the translated text is written directly into the same zones. All images, vector diagrams, logos, and background elements remain 100% untouched.
- Accurate DOCX (pdf2docx + substitution): For editing in Microsoft Word, we create a `.docx` based on the original PDF using the `pdf2docx` engine, then substitute translated phrases into the already-formed native tables and blocks. This guarantees a much better layout than standard text-flow exports.
- Alternative formats (DOC, HTML, TXT):
- .DOC: A fallback path for compatibility with older office suite versions.
- .HTML: For quick viewing in a browser or publishing on the web.
- .TXT: For getting clean translated text without formatting.
4. Convenient UX: Progress, Preview, and Local History
We made the translation process as transparent and comfortable as possible:
- Real-time tracking: A progress indicator shows which batch of text is currently being processed by the neural network.
- Interactive preview: You can quickly evaluate the result in PDF or HTML format right in the browser window before downloading.
- History without page reload (IndexedDB): All translated documents and their versions are instantly saved in your browser's local storage. You can return to them at any time, even after closing the tab.
5. Use cases: who is this service for?
The service is ideal for professionals who work with complex foreign documents every day:
| Document type | Translation and preservation specifics |
|---|---|
| Technical manuals and instructions | Preservation of drawings, diagrams, instructions, and element labels in their original positions. |
| Scientific articles and publications | Preservation of multi-column layout, data tables, reference lists, and charts. |
| Analytical reports and presentations | Accurate translation of tabular arrays, diagrams, and financial indicators. |
| International contracts and agreements | Preservation of parallel text structure, bilingual tables, and legal details. |
6. Technical limits and realistic expectations
To ensure high speed and processing quality, the following parameters apply:
- File limits: Document volume up to 20 pages, size up to 25 MB.
- File requirement: Only digital PDFs with a text layer are translated. If the file is a scanned paper document, first run it through optical character recognition (OCR).
- Complex and web layouts: Documents with very dense typesetting, specific web designs, or multi-level columns may require minor manual adjustments in Word after translation due to differences in phrase lengths between languages.
- Terminology control: Despite Gemini AI's high accuracy, key highly specialized terms, proper names, and abbreviations in critical documents should be additionally verified after translation.
7. Data security and privacy
We minimized data transfer to third-party servers to protect your confidentiality:
- Only structured text split into fragments is sent to the AI side.
- The original PDF file is sent to our server once — only at the final assembly stage of the resulting document.
- Immediately after generating the finished file, all temporary materials are permanently deleted from the server.
- Your translation history is stored exclusively in your browser and is not linked to accounts or external databases.
Frequently asked questions
How many pages can be translated at once?
Up to 20 pages and up to 25 MB per file. Split a larger document into parts. That makes both the translation and the layout assembly more stable.
Will images, diagrams, and tables stay in place?
In PDF format, yes, as far as the layout allows: images and lines stay from the original, and mostly the text in the same areas is replaced. In DOCX, the structure is rebuilt for Word (tables and blocks), but very intricate layouts may shift.
Which format should I choose: PDF, DOCX, or TXT?
PDF when the look "as in the file" matters. DOCX when you will keep editing in Word. DOC is the older Word format. HTML is convenient for viewing in a browser. TXT is text only, without formatting.
Does the service see my entire PDF during translation?
The text for translation is extracted in your browser; only structured text goes to the AI, not "extra" data from your disk. The original PDF is sent to the server once, after translation, only to assemble the chosen format, and is immediately deleted from temporary files.
Why isn't the document translated all at once?
Long texts are split into parts due to language model limits. This is normal: translation chunks come first, then a single finished file is assembled.
What should I do with a scanned PDF?
If the text can't be selected with the mouse, it's images of pages. First run OCR (a separate tool on filltools.com), get a PDF with a text layer, and translate that here.
Where is the translation history stored?
Only in your browser on this device. No account login is needed; after clearing site data or on another computer, the history will be gone.
Reviews (0)
Similar tools
Other tools from this category you may need next.
Image to PDF Converter
Convert images (PNG, JPEG, WEBP) into a PDF document. Each image becomes a separate page.
PDF to Word Converter
Online PDF to Word (.docx) converter running server-side via pdf2docx: preserves tables and page st…
PDF Compressor
Compress PDF files directly in your browser without uploading them to a server. Uses the pdf-lib li…
