Scanned PDFs in OmegaT, MateCat and Wordfast: What Each Can Do

A client sends a scanned birth certificate as a PDF, the deadline is Friday, and your CAT tool is a free or cheap one. You drop the file into OmegaT and get zero segments, or MateCat warns that OCR for your Cyrillic source is experimental, or Wordfast Pro’s converter finds nothing to convert. If you need to translate a scanned PDF in OmegaT, MateCat or Wordfast, the answer differs per tool, and only two of the five products involved read scans at all.

Below: one answer table, the steps per tool, the Word-first route that works with all of them, a pre-import checklist and the documented pitfalls. General reviews of these tools live in the OmegaT guide for beginners and the free cloud CAT tools comparison; this article is only about scans.

What OmegaT, MateCat and Wordfast do with a scanned PDF¶

A scanned PDF is a picture of a page inside a PDF container. Before a CAT tool can segment it, something has to run OCR (optical character recognition, the step that turns letters in an image into text). Here is who does that, as of October 2026:

Sources for each row are linked in the tool sections. Two points stand out: OmegaT and Wordfast Pro need the OCR done elsewhere, and the two products that read scans (MateCat and Wordfast Anywhere) are both online, which matters if your client’s NDA forbids cloud processing.

First check whether your PDF is really a scan: try to select a line and paste it into a plain text editor. If nothing pastes, or the whole page highlights as one block, it’s image-only. The longer version of the test, including mixed files, is in why Trados and memoQ fail on scanned PDF import; the logic is the same here.

Why a scan breaks in a CAT tool¶

A CAT tool works on text: it splits sentences into segments, matches them against your translation memory and writes the translation back into the same structure. A scan gives it pixels. OCR has to read the page first, and then a second job has to rebuild the layout (paragraphs, tables, columns) in a format the CAT tool understands. Most of the trouble comes from that second job.

OmegaT shows the mechanism clearly because its code is public. According to OmegaT’s PDF filter source, the filter uses Apache PDFBox to strip the existing text layer sorted by position on the page, joins lines into paragraphs until it meets a blank line, and writes the target as .txt. A scan has no text layer, so there is nothing to strip.

The How to chapter of the OmegaT manual puts PDF in a category of its own:

Some formats, like PDF, cannot be properly handled through “round trip” conversions. They require an intermediate conversion to a supported format that serves as a base to manually create a proper target language document.

So a PDF is never a source you translate and send straight back; you translate an editable copy and rebuild the document by hand. The same paragraph warns that the conversion “will always require extensive manual adjustments”.

Scans add their own problems. Stamps overlap printed text, signatures cross lines, handwritten dates read badly, and a QR code is a black square to a text recognizer. Tables lose their reading order when OCR reads straight across two columns (more in why CAT tools break PDF tables and multi-column layouts).

Rebuilding a scan into a usable Word copy is a job we know from building it. In ChatsControl, a vision model looks at each page and rebuilds it as a Word copy, then a second check compares that copy with the page line by line before anything is translated. Each scan gets a quality grade (good, fair or poor), because handwriting and very degraded scans are still read unreliably.

OmegaT: no OCR, so the scan becomes a Word file first¶

OmegaT 6.1 (6.1.0 released on 2 August 2026, per the OmegaT changelog) has no OCR or image features. The PDF input filter arrived in OmegaT 3.0.1 on 13 May 2013 and still extracts only text that already exists. So for a scan, OmegaT has no route of its own; you feed it a file it handles well.

1. Test the PDF. If text selects and pastes, the PDF filter will extract it. If not, go to step 3.
2. For a text PDF, decide whether .txt is acceptable. Put the PDF in the project’s source folder, translate, and run Project > Create Translated Files (Ctrl+D). The target lands as name.txt with no formatting: fine for a gist, useless for a certificate.
3. For a scan, OCR it into DOCX outside OmegaT and check the Word file against the scan.
4. Clean the DOCX (see the cleanup list below).
5. Put the DOCX in the source folder. OmegaT handles DOCX through its Office filter and keeps formatting as tags, so the translated Word file keeps the layout of your cleaned source.
6. If the tags are unbearable, turn on Hide tags in Project > Properties (Ctrl+E).

Hide tags exists for exactly this case. The Project Properties section of the OmegaT manual describes it like this:

This is especially useful when dealing with texts where inline formatting is not particularly useful (e.g., OCRed PDF, or poorly converted .odt or .docx files.)

The trade-off follows right below: hidden tags are stacked at the end of the segment, and you apply bold, italics and other formatting by hand in the translated file. For Word 2007+ files, the same page points to the Tagwipe script if all you want is fewer tags.

One more OmegaT warning: settle the final DOCX before you start. The manual cautions that changing files in the source folder partway through a project can make segments disappear or new ones appear. Found an OCR error on page 3 mid-project? Fix it in the target and correct the source DOCX after delivery.

MateCat: OCR on upload, DOCX out¶

MateCat reads the scan as part of the upload, with no separate conversion step, and it costs nothing. As MateCat’s own site puts it:

It’s free for translation companies, translators and enterprise users.

MateCat’s supported formats guide lists “Scanned files (PDF and images)” among its 80 file formats and claims more than 200 languages (as of October 2026). PDF files “will be downloaded as docx in order to be still editable”, and for PDFs exported from DTP software the guide suggests uploading the original file instead. According to MateCat’s open-source code, the accepted image types are bmp, png, gif, jpeg, jpg, jfif and tiff; the hosted service may run different settings.

The steps:

1. Upload the scan (PDF or image) on the MateCat home page and pick the language pair.
2. Read the upload message. For non-Latin scripts the code shows a warning (quoted below) telling you to download the preview and check the conversion. Do it before translating a single segment.
3. Translate in the editor as usual, with TM and MT.
4. Download the result. At 100%, the Draft button in the Download menu turns into Download translation.
5. Compare the DOCX with the scan. Per the code, .docx is appended to the full original name (scan.pdf becomes scan.pdf.docx).

MateCat’s download guide explains why step 5 matters:

There is an exception with source files in PDF or JPG. In this case, once translated, it will be downloaded a Word DOCX target file.

The next sentence adds that for PDFs and scanned images “the output could have a similar but not identical layout and some manual editing could be required.” MateCat gives you a Word file to fix, not a finished certificate. The same guide says a Draft download replaces unedited parts with machine translation, so never send a draft by mistake.

For Ukrainian, Russian, Greek or any other non-Latin source, the upload warning matters most. The text below sits in MateCat’s source code as of 13 October 2026:

File uploaded successfully. Before translating, download the preview to check the conversion. OCR support for non-latin scripts is experimental.

The code also holds an error saying OCR for right-to-left languages isn’t supported, while its language data flags Arabic and Hebrew as OCR-capable. What the hosted service does with an Arabic scan today is unclear, so test one page before you quote.

Wordfast Pro, Anywhere and Classic: three products, three answers¶

Wordfast sells three translation tools: Pro reads only text PDFs, Anywhere converts scans, Classic relies on Word.

Wordfast Pro: no OCR, convert first¶

The Wordfast Pro 11 help lists “Portable Document Format (*.pdf)” as supported and says Live Preview works with DOCX files converted from PDFs, so Pro converts PDFs to DOCX internally. The help never mentions OCR, scanned PDFs or any image format.

Wordfast’s wiki page on translating PDF files with Wordfast Pro (last edited 14 April 2021) is direct: the bundled converter handles only “live” PDFs (selectable text) and tries to keep layout, formatting and fonts. For “dead” PDFs (scans and photos) it tells you to convert the file first with Wordfast Anywhere. So in Pro you convert the scan to Word in Anywhere or an OCR program, check and clean it, translate the DOCX, and export DOCX.

Wordfast Anywhere: scan in, Word file out, you check it¶

Wordfast Anywhere (WFA) is the online Wordfast and the product Wordfast points to for scans. The Wordfast Anywhere help (FAQ 4.2.9) describes the flow:

1. Upload the PDF like any other file. The upload list has “pdf” and “tif tiff” but no JPG or PNG, so turn a phone photo into a PDF first.
2. Follow the disclaimer and download the Word version with the button at its bottom.
3. Review the Word document against the scan.
4. Upload the Word version and translate it.
5. Download the translation; the help says a PDF download is available too.

The help warns that the conversion “may take some time (up to 10 minutes) depending on the size and content of the PDF”. Password-protected or permission-restricted PDFs can’t be converted at all, and the maximum single upload is 200 MB, zip files accepted (as of October 2026, per the same help).

Per Wordfast’s Anywhere page, WFA was free from 2010 to 2022 and became paid on 1 January 2023: $9.90 a month, a one-month trial for $1, and free while you hold a valid Wordfast Classic or Pro licence. Without credit, the WFA help says, an account drops to Limited mode: 15,000 translated or edited words a month, with “powerful document conversions” still listed as included. Hit the cap and the account moves to Basic mode until month end, where uploads are blocked.

The help names no OCR engine, recognition languages or page limits, so test a page in your language pair before quoting.

Wordfast Classic: whatever Word gives you¶

Wordfast Classic is a Word add-on, so it translates only what Word can open. The WFC 9 manual says that “starting in 2019” Word opens PDFs as editable documents (the manual’s framing; the feature itself is older). For dead PDFs, it advises keeping a Wordfast Anywhere subscription for its scanned-PDF conversion.

Classic’s PDF/Web Text Rebuild function (Tools > Misc in the main WFC setup dialog) tries to rejoin paragraphs that a conversion broke with a hard return at every line end. It also flattens odd font sizes and merges skinny columns, and the manual reminds you to “always process a copy of a document!” The same manual is frank about PDFs as a source:

Note that no tool will let you translate a PDF “from within”, and deliver a visually correct translated PDF at the push of a button. PDFs must be converted to another format, and even that process is hazardous.

Read that as a pricing note too. The manual’s disclaimer lists “PDF conversions, OCRed material” among documents that need expert handling, so if a scan must come back looking like the original, the layout work belongs in the quote (how to quote a scanned document when word count finds nothing).

The route that works in all three: a checked Word file first¶

Whatever tool you use, the most controllable route is the same: get a clean, checked DOCX out of the scan, then import it like any client Word file. OmegaT, MateCat and all three Wordfast products handle DOCX as a first-class format.

With an OCR program¶

Word, Adobe Acrobat, ABBYY FineReader and Google Docs all turn scans into editable text, with different results on tables and forms; the comparison is in scanned PDF to editable Word before translation. On top of any of them:

1. Fix the image first: straighten, rotate, rescan if you can (what to fix first in a poor-quality scan).
2. Set the OCR language to the document’s real languages.
3. Export to DOCX, not RTF or TXT, so tables stay tables.
4. Compare the DOCX with the scan: names, dates, ID numbers, amounts.
5. Clean it up (next list), then import and look at segments and tags before translating.

Cleanup before import, for any OCR output:

• Line breaks inside sentences. A hard return after each printed line makes every fragment its own segment. Join them (Wordfast Classic’s Text Rebuild, or a careful find-and-replace on paragraph marks in Word).
• Text boxes and frames. Floating boxes make the segment order differ from the page order, and a box the filter misses stays untranslated, so turn them into paragraphs or table cells.
• Reading order. A two-column page should read column by column, not straight across.
• Font and spacing noise. OCR output carries many small font, size and spacing changes, and each can become a tag. Clear direct formatting on body text, or use Hide tags in OmegaT.
• Stamps and signatures. Decide how to render them, for example as bracketed notes (translating stamps and seals).

With a scan-to-Word service¶

One option is ChatsControl: you upload the scan or photo (PDF, PNG, JPG, TIFF, WebP), it reads the page and returns an editable Word copy that follows the page layout. Text is typed as printed, tables come as real tables, stamps, seals, signatures and photos are cut from the page and placed where they stand, and QR codes are redrawn clean. You then import that .docx into OmegaT, MateCat or Wordfast like any client Word file (or have it translated directly in the same service).

1. Upload the scan or photo and wait for the job to finish.
2. On the job page, download the Word copy of the original with the Original button (next to the translated Word file and the raw scan).
3. Check that copy against the scan: names, numbers, dates and the quality grade.
4. Import the source-language .docx into your CAT tool and translate with your own TM and glossary.

The result is an ordinary Word document; OmegaT, MateCat and Wordfast import it the way they import any client .docx. Check the tag view after import, as with any converted file. To see the Word copies first, browse real scans next to the Word files the service returned.

Prices as of October 2026, per the ChatsControl app: a free tier of 6,000 characters a month, then pay as you go at $1.50 per 1,000 characters ($2.70 per 1,800-character page). Scans are billed by recognised characters, at least 1,000 characters per page. Each job also produces a translation and is billed as above; there’s no separate OCR-only price. Uploads are deleted automatically after 30 days by default.

The limits:

• ChatsControl isn’t a CAT tool or a TMS: no translation memory, no segment-level project management.
• It’s cloud only, with no on-premise or offline mode.
• Handwriting and very degraded scans are read unreliably, and the quality grade says so (handwritten documents and OCR).
• Stamps, seals and signatures stay as pictures where they stand. If the authority wants the seal’s wording rendered, you add a bracketed note such as [Seal: …] yourself; the stamp’s text isn’t translated inside the picture.

The routes compared, and a pre-import checklist¶

No route skips the check against the scan.

Pre-import checklist (copy it into your job notes):

What goes wrong with scans in these tools¶

OmegaT returns plain text for a PDF. People expect a translated PDF and find name.txt in the target folder. Per the PdfFilter source, that’s by design. Use the PDF filter for reference reading, never for a layout job.

Password-protected PDFs. OmegaT’s changelog lists a fix in 4.1.0 (January 2017) for “Password-protected PDF prevents project from loading”; the project now loads, but the filter still throws an error on that file. Wordfast Anywhere can’t convert protected PDFs either. Ask the client for an unlocked file first.

Broken paragraphs from hard returns. A paragraph mark after every printed line turns each line into a segment, so TM matches drop and the translation reads in fragments. Wordfast Classic’s manual has a tool and a separate macro just for this.

MateCat’s file name and draft. The target comes back as scan.pdf.docx, so rename it before delivery. A Draft download fills unfinished segments with MT, per MateCat’s download guide.

Non-Latin scans in MateCat. OCR for non-Latin scripts is labelled experimental. For a Ukrainian or Russian certificate, check the preview before you commit to a deadline.

OCR programs and certificates. Certificates were a weak spot for OCR long before AI. In a 2010 ProZ thread on PDF and JPG files in OmegaT, Susan Welsh described her results with ABBYY like this:

For a certificate, or anything with lots of graphics and tables and text boxes, I have not found it satisfactory.

OCR has improved since September 2010, but certificates still mix stamps, tables, small print and handwriting on one page. Another poster in that thread got good results by marking text and graphics areas by hand instead of trusting automatic layout detection, at about a minute per page.

Delivery. None of these tools hands you the scan’s layout as a PDF. OmegaT gives .txt (or DOCX for a DOCX source), MateCat and Wordfast give Word files, and the Wordfast Anywhere help offers a PDF download without saying how close it looks to the scan. If the client wants a PDF, save the checked Word file as PDF and budget the layout time. After translation, the free translation QA check compares the file with the source for numbers and meaning, which catches digits OCR got wrong upstream.

Agencies weighing these tools for scan-heavy work can compare them in memoQ vs Smartcat vs Wordfast for a small agency and see the paid-tool version of this problem in the memoQ scanned PDF guide.

FAQ¶

Can OmegaT translate a scanned PDF?¶

Not directly. OmegaT has no OCR, and its PDF filter only reads an existing text layer, writing the result as plain .txt. OCR the scan into a Word file first, check it, and put the DOCX into the project’s source folder.

Does MateCat do OCR on scanned PDFs and images?¶

Yes. MateCat’s guides list scanned PDFs and images as supported, and a PDF or JPG source is downloaded as a Word DOCX. Its open-source code calls OCR for non-Latin scripts experimental, so download the preview and check it before translating.

How do I translate a scanned PDF in Wordfast Pro?¶

Convert it outside Wordfast Pro first. Pro’s converter only handles PDFs with selectable text, and Wordfast’s wiki points scanned PDFs to Wordfast Anywhere. Download the Word version there, check it, and open that DOCX in Wordfast Pro.

Is Wordfast Anywhere free for converting scanned PDFs?¶

Not since 1 January 2023. It costs $9.90 a month, with a one-month trial for $1, and it’s free while you hold a valid Wordfast Classic or Pro licence (as of October 2026). Limited mode keeps document conversions but caps you at 15,000 words a month.

Will the tags and segments be messy after OCR?¶

OCR output can make them messy. Hard returns after every printed line split sentences into fragments, and font or spacing changes turn into extra tags. Clean the DOCX before import, use OmegaT’s Hide tags option for OCR output, and check the tag view after import, as with any converted file.

Is it safe to upload client documents to MateCat or an online OCR?¶

Check the client’s confidentiality terms first. MateCat, Wordfast Anywhere and any online OCR process the file on someone else’s server; OmegaT, Wordfast Pro and an installed OCR program keep it on your machine. If the contract forbids cloud processing, stay offline.

Can I deliver a translated PDF back to the client?¶

Not without extra work. OmegaT returns .txt for a PDF source, MateCat returns DOCX, and the Wordfast Classic manual says no tool translates a PDF from within and returns a visually correct one. Deliver the checked DOCX, or save it as PDF after layout work.

Sources¶

1. OmegaT changelog (changes.txt)
2. OmegaT PdfFilter source code
3. OmegaT manual: How to…
4. OmegaT manual: Windows and Dialogs
5. MateCat: About
6. MateCat guides: Supported formats and languages
7. MateCat guides: Download your translation
8. Wordfast Pro 11 help
9. Translating PDF Files with Wordfast Pro (Wordfast wiki)
10. Wordfast Anywhere help
11. Wordfast Anywhere pricing and access
12. Wordfast Classic 9 manual
13. ProZ forum, OmegaT support: working with PDF and JPG files (2010, archived)