OmniDocTranslate Scanned Chinese and Japanese PDFs to English
Learn how to translate a scanned Chinese or Japanese PDF into English, with checks for vertical text, character recognition errors and bilingual output.
If you have a scanned Chinese or Japanese PDF, such as a contract, a government notice, a product manual or a research paper, you probably can't search it, copy from it or read it in English without help. Scans are pictures of pages, so the text has to be recognized before it can be translated. With Chinese and Japanese, that recognition step has extra failure points: dense characters, mixed scripts, small annotations and vertical columns that many tools read in the wrong order.
This guide is for anyone who needs to translate a scanned CJK document into English and wants a usable result. You'll learn how to tell whether a scan is a good candidate, what to check before you upload, how to run the translation step by step, and how to catch the problems that are specific to Chinese and Japanese layouts. The method works the same way for both languages, with a few differences noted below.
How do you translate a scanned Chinese or Japanese PDF to English? The short answer
Upload the scanned PDF to OmniDoc, choose English as the target language, review a bilingual copy against the original, and fix any recognition or reading-order errors before you rely on the English text.
| Your situation | What to do | Why |
|---|---|---|
| You only need to read the English | Translate the scan directly and choose translation-only output | This is a quick route when you won't edit or quote the text |
| You need to check the translation against the original | Choose bilingual side-by-side output | Each source passage sits beside its English version, so errors are easy to spot |
| The page has vertical columns or small annotations | Check a few pages for reading order before translating the whole file | Vertical text and furigana are common places where the reading order goes wrong |
| You need to edit, compare or quote the source text | Run OCR in the desktop client first, then work from the searchable copy | A searchable PDF lets you select, copy and compare text reliably |
| The scan is blurry or skewed | Re-scan the pages at higher quality before translating | Poor images cause recognition errors that appear as wrong characters in the English |
Before you start
Spend a few minutes on these checks. They take less time than fixing a bad translation later.
- Confirm the file is actually a scan. Open the PDF and try to select a line of text. If you can't select it, or it looks like a picture of a page, it's a scan. A PDF that already has selectable text is a different case and may not need recognition at all.
- Look at the layout on several pages. Check whether the text runs in vertical columns, mixes horizontal and vertical blocks, or includes stamps, handwritten notes or tables. Mixed layouts are the usual reason a translation reads out of order.
- Note the script and the variant. Chinese scans may use simplified or traditional characters. Japanese scans mix kanji, hiragana and katakana, often with small furigana readings above the kanji. Knowing which variant you have helps you judge whether the recognized text looks right.
- Decide what the English is for. A quick reading copy can use translation-only output. A document you'll cite, sign or send to someone else needs bilingual output and a proofreading pass.
- Keep the original. Save an untouched copy of the scan before you run OCR or edit anything. If a later step introduces an error, you can compare against the source.
Translating the whole PDF in OmniDoc, step by step
- Pick one sample page to test. Choose a page with dense text or a layered layout, such as a vertical column with an annotation. Use it to judge the output before you commit to the full file.
- Decide whether you need OCR first. If you only want to read the English, you can upload the scanned PDF as it is, since scanned PDFs are supported for translation. If you plan to edit, search or compare the source text, run OCR in the desktop client's PDF editor first and keep the searchable copy.
- Upload the PDF on omnidoc.top or in the desktop client. You can start from the download at download.omnidoc.top if you prefer to work on your computer. Set the source as Chinese or Japanese and the target as English so the translation matches the document's language.
- Choose the output format. Select bilingual side-by-side output while you check the work, because each original passage sits next to its English version. Switch to translation-only output once you trust the result and only need to read it.
- Pick a style that matches the document. OmniDoc offers style options for academic, legal, medical, technical, business and literary text, and it can keep terms consistent across the file. Choose the closest match, since a contract and a product manual use different vocabulary.
- Download the result and open it beside the original. Check the layout, tables, formulas and pictures, which OmniDoc keeps during translation, and compare the opening pages of text with the scan.
- Edit errors directly if needed. The desktop PDF editor can change paragraph text in place. For longer corrections, export to Word, edit there, and save the final version under a new name so you keep the PDF output intact.
- Re-run the translation for problem pages only. If a few pages came out badly, fix the source scan or the OCR for those pages and translate them again, rather than reprocessing the whole document.
Check the result
Machine translation and text recognition can both make mistakes, so check the important parts yourself before you use the English for anything that matters. Start with the places where CJK documents tend to fail.
| Check | How to see it | If it's wrong |
|---|---|---|
| Reading order | Compare a vertical column with its English paragraph. Do the sentences follow the original sequence? | Run OCR in the desktop client, then translate the searchable copy. If the order is still wrong, split the page or re-scan it with the page held straight. |
| Recognized characters | Pick out a few kanji, traditional or simplified characters, and numbers in the bilingual view. Do they match the scan? | Re-scan at higher quality. Blurry strokes often turn into similar-looking characters. |
| Furigana and small annotations | Look for readings or notes above or beside kanji. Are they translated as extra text or dropped? | Decide whether you need them. If they matter, check those passages against the original by hand. |
| Names, product terms and figures | Compare every name, model code, date, amount and reference number with the source. | Correct them in the document and re-check the surrounding sentences, since one wrong term can change a whole paragraph. |
| Layout | Check tables, headers, footers and page numbers against the original. | Use the bilingual output to locate the problem, then fix the affected page in the editor. |
For legal, medical or financial documents, have a fluent reader or a qualified professional check the key passages before anyone acts on them.
🎯 Which approach fits you?
The right approach depends on what you'll do with the English. A reader who only needs the gist of a Japanese notice can work from translation-only output and skip OCR. A researcher quoting a Chinese paper needs OCR, bilingual output and a careful check of every quoted passage. Someone editing a contract for a negotiation should export to Word, make corrections there and keep the bilingual version as the reference copy.
❓ FAQ
1. Can I translate a scanned PDF without running OCR first?
Yes. Scanned PDFs can be uploaded for translation as they are. Run OCR in the desktop client first only if you need selectable, searchable text for editing, comparing or quoting.
2. Why does the English come out in the wrong order on some pages?
Vertical Japanese and Chinese columns, mixed layouts and annotations can confuse the reading order. Check those pages against the original, then try OCR on the searchable copy or re-scan the page straight. If it still fails, split the page and translate the sections separately.
3. Should I use bilingual or translation-only output?
Use bilingual output while you are checking, since it shows each source passage next to its English version. Use translation-only output for a final copy you only need to read.
4. Does the translation keep the original layout?
OmniDoc keeps layout, tables, formulas and pictures during translation. Still compare pages with the original, because a poor scan can cause problems that look like layout errors.
📝 Summary
Translating a scanned Chinese or Japanese PDF into English works well when you check the text recognition and reading order before you trust the output. Test a sample page, choose bilingual output while you review, and fix problem pages at the source. For any document you will act on or cite, have the key passages checked by a fluent reader.
