by Jessie Doherty, active member of the LTD’s Leadership Council
Every translator has met that PDF.
Maybe it is a crooked scan of a birth certificate. Maybe it contains faded stamps, handwritten notes, tables, seals, signature blocks, and formatting apparently designed to test both your desktop publishing skills and your patience. You know you can translate the words. The bigger question is: How long will it take to rebuild the document?
That was the problem at the heart of the October 7th ATA TEKTalk, “EZDocuAI Versus Your Ugliest PDF—OCR, Layout, and Translation.” The session demonstrated a tool designed not simply to translate documents, but to tackle one of the less glamorous—and often more time-consuming—parts of a translator’s job: turning difficult source documents into usable, editable files.
Meet EZDocuAI—and the Man Behind It
Roberto Rodríguez, CEO and co-founder of EZDocuAI, is a mechanical engineer and entrepreneur originally from Barranquilla, Colombia. After immigrating to the United States, he built a career combining artificial intelligence and business consulting. Along the way, he noticed something: professional translators were essential to global communication, yet their very specific workflow problems were often overlooked by technology developers.
Rodríguez and his co-founder set out to address that gap, spending months learning how translators actually work and where they lose time. EZDocuAI grew out of that effort, with a particular emphasis on desktop publishing (DTP) and difficult documents.
The philosophy Rodríguez emphasized throughout the webinar was equally important: technology should serve professionals, not replace them.
That philosophy became much clearer during the live demonstration.
EZDocuAI is browser-based and designed for both individual professional translators and companies managing teams of translators. Its dashboard includes document uploading, team management, project statistics, terminology management, translation memory, a knowledge base for reference documents, and templates. Most importantly for anyone who has ever stared unhappily at a barely legible certificate, it is designed to handle scanned and handwritten documents.
From OCR to Translation: What EZDocuAI Actually Does
During the webinar, Rodríguez demonstrated the workflow using a scanned birth certificate.
The original scan appeared on one side of the workspace while project information—including languages and an existing glossary—could be managed alongside it. EZDocuAI analyzed the document, extracted its contents, and reconstructed the layout. But it did not simply announce, “Voilà!” and send the machine’s work out the door.
Instead, the professional remained involved throughout the process.
The system presents questions and items for verification, allowing the translator to check the source, terminology, translation, and formatting. The reconstructed document can also be edited during the DTP stage. From there, the translator can download the editable source and continue working in another CAT tool, or remain in EZDocuAI and proceed with translation. Final output can be exported as PDF or DOCX, including bilingual versions and a Certificate of Accuracy.
That human-in-the-loop approach may be the most important feature of the platform. EZDocuAI’s AI-generated translation is intended as a professional draft—not a finished translation that bypasses human judgment. Translators review, correct, approve, and ultimately take responsibility for the final product.
The OCR and layout capabilities are particularly interesting. EZDocuAI can process digital files, scans, handwriting, complex tables, government forms, seals, stamps, and multi-column layouts. Users can choose an output that closely replicates the original layout, combines translated text with original visual elements, or produces a more editable document using conventional paragraphs, headings, and tables.
Terminology tools are built into the workflow as well. Glossaries can be created manually, imported from Excel or CSV, or extracted from previous source-and-translation pairs. Teams can share terminology, while translation memories can be exchanged with tools such as Trados through TMX files. There is currently no direct Trados plugin.
There are limitations. Parallel bilingual source documents can produce duplicated translations, for example, and two photographs of different portions of the same page are not automatically stitched together. There is also no standalone word-count function for quoting a job without creating an order. In other words, the ugly PDF has not yet been completely defeated.
What does it cost?
EZDocuAI uses a credit system: one credit equals one processed page, regardless of whether that page contains ordinary digital text, requires OCR, or includes handwriting requiring AI transcription.
Users can buy non-expiring pay-as-you-go credit packages, subscribe to monthly or annual plans with credits that replenish each billing period, or enable automatic recharging. Subscription tiers such as Starter, Team, and Enterprise can add features including team seats, custom certificate branding, additional glossaries, and API access. The webinar also announced a seven-day or 15-credit free trial and a 20% discount for ATA members.
For confidentiality, the webinar explained that documents are encrypted and restricted to the translator and authorized team members. The follow-up information also states that customer documents and translated segments are not used to train AI models. Files can be deleted from the server after translation.
The Real Goal: Spend More Time Translating
Perhaps the best way to understand EZDocuAI is not as another AI translator, but as an attempt to reduce everything surrounding the translation.
During the webinar, Rodríguez estimated processing from OCR toward translation at roughly three to ten minutes, with a typical birth certificate taking around three minutes per page before professional verification. The point is not to eliminate the translator. It is to eliminate—or at least substantially reduce—the hours spent reconstructing tables, repositioning fields, deciphering scans, and wrestling a PDF into submission.
For translators, that distinction matters. AI may be doing more of the mechanical work, but professional judgment remains where it belongs: with the human.
If you are heading to ATA67 in San Francisco, October 28–31, you will have another opportunity to hear from the EZDocuAI team: Rodríguez confirmed during the TEKTalk that they will be presenting at ATA67.
So if your professional life occasionally includes a PDF that makes you wonder what you did to deserve it, EZDocuAI may be worth a closer look.
The ugly PDFs, after all, are probably not going away.
The tools we use to deal with them can certainly get better.
Q&A
Q: How many languages does EZ Docu AI support? Could you provide a full list?
EZDocuAI currently supports 101 languages:
Afrikaans, Albanian, Arabic, Aragonese, Armenian, Assamese, Aymara, Azerbaijani, Bashkir, Basque, Belarusian, Bengali, Bosnian, Breton, Bulgarian, Burmese, Catalan, Chinese, Croatian, Czech, Danish, Dutch, English, Esperanto, Estonian, Filipino (Tagalog), Finnish, French, Galician, Georgian, German, Greek, Guarani, Gujarati, Haitian Creole, Hausa, Hebrew, Hindi, Hungarian, Icelandic, Igbo, Indonesian, Irish, Italian, Japanese, Javanese, Kazakh, Korean, Kyrgyz, Latin, Latvian, Lingala, Lithuanian, Luxembourgish, Macedonian, Malagasy, Malay, Malayalam, Maltese, Maori, Marathi, Mongolian, Nepali, Norwegian, Occitan, Oromo, Pashto, Persian, Polish, Portuguese, Punjabi, Quechua, Romanian, Russian, Sanskrit, Serbian, Sesotho, Slovak, Slovenian, Spanish, Sundanese, Swahili, Swedish, Tajik, Tamil, Tatar, Telugu, Thai, Tsonga, Tswana, Turkish, Turkmen, Ukrainian, Urdu, Uzbek, Vietnamese, Welsh, Wolof, Xhosa, Yiddish, Zulu.
If you need a language that is not on this list, contact our support team and we will evaluate it.
Q: What is the best way to learn EZDocuAI? Any seminars? User documentation? Coaching sessions?
EZDocuAI is designed to be learned by doing: the workflow guides you step by step from upload to review and export.
If you have a question at any point, the AI assistant built into the platform can guide you through any feature. When there is a tutorial video or other material that can help, the assistant will point you to it directly.
You can also watch our tutorial videos on our YouTube channel: youtube.com/@EZDocuAI
We are also expanding our learning resources. Coming soon: a “What’s new” section inside the platform, live webinars that walk through specific features, and a refreshed newsletter for registered users with updates and tips.
Q: How are glossaries created and added?
Glossaries are managed in the My Terms section. We recommend organizing your terminology into folders by specialty (for example, Legal, Medical or Civil Registry). Inside each folder, you can create as many terminology sets as you need.
When you create a terminology set, you can import your existing resources directly: translation memories exported from your CAT tools, as well as Excel or CSV files.
To help EZDocuAI reflect your personal translation style, we also recommend two additional steps:
- Add a reference pair: upload an original document together with your own translation of it, so the platform can learn your style and terminology choices.
- Use the Knowledge Base: add reference articles or materials related to your specialty to further improve the quality of the output.
Q: Does the tool support Nepali? Does the tool support Arabic?
Yes, both Nepali and Arabic are among the 101 languages EZDocuAI supports.
Q: For example in former USSR countries a lot of documents are in bilingual format at the same time: in Russian and let’s say Uzbek, Tajik, Kazakh or Armenian, etc. I ran a test and the translation delivered duplicated the translation, so the English output was doubled. Can this be adjusted somehow, and if yes, would the adjustment to be done occur on the user side or on the company side?
Thank you for testing this case. Your feedback is exactly the kind of real-world input that shapes the product.
This is now handled on the user side. We have just released a new feature: when EZDocuAI detects more than one language in a document, it asks you how you want to proceed. For example, you can translate both language versions and merge them into a single output, or take another path that you consider appropriate as a professional.
It takes just a couple of clicks, and you stay in control of the decision, as the translator should.
If you run your test document again, you will see the new options.
Q: Is it compatible with US vital records with complex tables and other formatting?
Yes. Complex official documents like these are where EZDocuAI really stands out.
Birth, marriage and death certificates and other US vital records are handled without issue. EZDocuAI reconstructs complex tables, multi-column structures, seals, stamps and registrar annotations, preserving the layout of the original.
Handwritten content is one of our key differentiators. Handwritten entries, notes and annotations are transcribed and translated alongside the printed text, so nothing is left out.
As always, you review and refine the result in the side-by-side editor before exporting your final document.
Q: Is the output format always in table format? What about if the source format is more complicated? Like a form?
No, the output is not limited to tables. EZDocuAI adapts to the actual structure of each document, whether it is a form, a narrative text, a certificate with tables, or a combination of all of them. The original layout is reconstructed, not forced into a single format.
of all of them. The original layout is reconstructed, not forced into a single format.
We encourage you to test the platform with your most complex documents. Real-world cases like yours are the best way for us to keep improving EZDocuAI and make it better for you every day.
Q: Can you provide more information on the pricing of the product?
EZDocuAI uses a simple credit-based model: 1 credit = 1 page. The rate is the same for every type of document. A digital PDF, a low-quality scan or a handwritten certificate all cost exactly 1 credit per page. Before you confirm any order, the platform shows you how many credits it will use, so there are no surprises.
We offer plans for every workload, from occasional freelance projects to large agencies:
| Plan | Price | Credits included | Extra credit |
| Free | $0 | 5 per month | $2.99 |
| Pro | $49/month | 25 | $1.99 |
| Max | $149/month | 100 | $1.49 |
| Team | $499/month | 400 | $1.25 |
| Business | $999/month | 1,000 | $0.99 |
| Enterprise | Custom | Custom | Custom |
All plans are also available annually with a 17% discount, which is the equivalent of two months free.
If you need more pages than your plan includes, you can buy additional credits at any time, so your work never stops in the middle of an urgent delivery.
For full details, visit the Pricing section on our website: ezdocu.ai
Q: What is the LLM in the background? Or, what is the background MT/LLM you use? Are your OCR and CAT engines based on any commercially available LLMs?
The specific models and architecture behind EZDocuAI are part of our proprietary technology, so we don’t disclose them. What we can share is how it works: EZDocuAI combines state-of-the-art AI technology with our own document processing pipeline, built in-house specifically for official and legal documents. That pipeline covers OCR, handwriting transcription, layout reconstruction and translation.
What matters most for your work is how your data is protected:
- Zero training: your documents and translations are never used to train any AI model.
- Zero retention: documents are deleted after export, and you receive a verifiable deletion certificate.
- Encryption: your documents are encrypted throughout the process.
We call this approach Zero Knowledge Processing: your clients’ information stays confidential, always.
Q: Considering how well it can read documents that other CAT tools can’t read, does it have a feature that only provides a word count for the document? If not, could this be added?
That’s a great use case, and a very practical one for quoting before committing to a full job.
Today, EZDocuAI provides the word count once a document has been processed. A standalone word count before processing is not available yet.
We have shared your suggestion with our product team. Since EZDocuAI can read documents that other CAT tools cannot, we agree this could be especially valuable for quoting complex scans and handwritten documents.
Q: If a client provides two images of one document, does this tool help merge the content into one full page?
EZDocuAI processes documents exactly as they are uploaded. Each page or image is processed as an individual page, uses 1 credit, and is delivered as its own page in the final document. If you upload three images, you’ll get three translated pages, using 3 credits.
If you want the translation delivered on a single page, combine the images into one image or one single-page PDF before uploading, using any image editor or PDF tool. EZDocuAI will then read the full layout as one page, and it will only use 1 credit.
Q: Does the output Word document formatting contain text boxes or frames with text boxes inside? Or is it generally formatted as tables and paragraphs?
It depends on the output option you choose. In the OCR Review step, EZDocuAI shows the available output options at the top of the screen, and you can switch between them to pick the one that best fits your needs:
- Option 1, text blocks: text is placed in independent blocks positioned to match the original layout. This is ideal when visual fidelity to the source document is the priority.
- Option 2, flowing text: text is delivered as standard flowing paragraphs and tables, without preserving the original images. This is the easiest format to edit in Word or to process in your CAT tool.
- Option 3, flowing text with images: combines flowing text with the original images, such as seals, stamps or logos. This option is available when the images in the source document are clear and not too large.
The options available can vary depending on the document.
Q: I’m new to official document translation. It looks like U.S. government agencies accept AI-translated documents. Is this the case, or do some agencies accept them and others don’t?
U.S. government agencies (such as USCIS, other federal agencies or state departments) do not accept raw, uncertified AI or machine translations on their own.
What they accept is a certified translation backed by a human translator. USCIS, for example, requires a certification in which the translator declares that they are competent to translate from the foreign language into English and that the translation is complete and accurate, along with their signature, full name, address and date.
Agencies do not evaluate whether a translator used CAT tools or AI to prepare the initial draft. Legal accountability rests with you, the professional: you verify every term, ensure accuracy and sign the certification under your own name.
That’s why EZDocuAI does not certify translations. The platform handles the heavy lifting (OCR, layout reconstruction, an AI-assisted draft and a customizable certification template), while you review the text, apply your professional judgment and issue the certification.
Q: How accurate is the AI translation provided by this tool? What is the average percentage of accuracy? I haven’t seen an AI tool that is 100% accurate.
You’re right: no AI translation tool is 100% accurate, and we don’t believe in advertising an accuracy percentage.
At EZDocuAI, the professional always has the final word. The platform is an assistant designed to boost your productivity: it takes care of the mechanical work (OCR, handwriting transcription, layout reconstruction and a first translation draft) so you can focus on what really requires your expertise.
We are constantly improving the accuracy of the platform, and there’s something even more valuable: EZDocuAI improves with you. When you work with a terminology set, it learns from your corrections every time you use it in your translations. Over time this reduces the number of edits you need to make. The more you use EZDocuAI, the better it fits your terminology, your style and the way you work.
These improvements happen entirely within your own private workflow. They stay in your account only: no other company or translator has access to them, and they are never shared or used to improve anyone else’s results.
Official documents involve critical details such as proper names, dates, numbers, registrar notes and jurisdiction-specific terminology. That’s why every output is treated as a working draft. The side-by-side review editor, your glossaries and your translation memories let you check every segment, make corrections and approve the final text before you deliver it.
Q: Is there any possibility of “purchasing” the software to run on a closed server for translators or offices who work with documents that cannot be shared or uploaded?
That’s a very valid concern for translators and offices handling sensitive documents. Deployment on private servers is something we are open to working on with you as part of an Enterprise plan. If this is a requirement for your operation, contact our team (hello@ezdocu.ai) and we will review the setup that best fits your needs.
In addition, EZDocuAI was built from the ground up for confidential documents through what we call Zero Knowledge Processing:
- Encryption: your documents are encrypted throughout the process.
- Zero retention: documents are deleted after export, and you receive a verifiable deletion certificate.
- Zero training: your documents and translations are never used to train any AI model.
Q: Can you connect this platform with Trados (to share your memories)?
There is currently no direct integration with Trados. However, our public API is scheduled for release by the end of October 2026. With it, you will be able to connect EZDocuAI to Trados through custom workflows, as well as to any other CAT tool, TMS or system you use in your operation.
In the meantime, you can already import your existing translation memories from your CAT tools into EZDocuAI through My Terms (see question 3).
Q: What are credits? Could you explain a bit more?
Credits are the unit EZDocuAI uses to process documents: 1 credit = 1 processed page.
The cost is the same no matter what the page contains. A digital PDF, a low-quality scan that requires OCR, or a handwritten document all use exactly 1 credit per page. Before you confirm any order, the platform shows you how many credits it will use.
Credits from paid plans and additional credit purchases never expire and accumulate month after month, so you never lose what you’ve paid for.
You can check your balance and transaction history at any time in the Credits section of the platform.