Translation quality, powered by AI.
Post-editing that scales with your business.
Translation volumes are growing. Deadlines are shrinking. Quality expectations never move. re[words] is built for exactly this: an AI-powered post-editing tool that analyses existing translations, fixes errors across six quality dimensions, and delivers corrected files — complete with a detailed quality report.
It’s not an autocorrect. It’s an intelligent quality layer — built by linguists, for linguists — designed for teams that work with XLIFF files every day.
re[words] is designed to reduce manual effort, not replace human judgement. For critical content, we recommend a final review by a qualified translator.
How it works.
Four steps. Built to fit your existing workflow without friction.
Drag and drop your XLIFF files directly into the interface. re[words] supports the most widely used formats: .sdlxliff (SDL Trados), .mxliff (Phrase / Memsource), .mqxliff (memoQ), and .xlf (Okapi Framework). Up to 20 files per job, with the option to attach a terminology glossary to guide the model’s output.
Define your language pair and select a processing mode: Automatic Post-Editing to correct existing translations, or LLM Translation to generate a first-pass translation from the source. Add a brief description of your content type to provide context for the model. For more specific needs, our team can build custom prompts tailored to your domain and register.
A recommended model comes pre-configured, selected based on independent quality evaluations. If needed, you can switch to OpenAI (GPT-4.1, GPT-5) or Anthropic (Claude Sonnet 4) depending on your content type and language pair. Your API key is used only for the duration of the job and never stored. You can also activate a Review pass with a different model to balance quality and cost on high-volume projects.
re[words] processes your files dimension by dimension — accuracy, fluency, style, terminology, locale, tags — and returns a corrected XLIFF file and an Excel quality report with scores, flagged issues, and applied fixes per segment.
Files are processed independently, so you can download results as soon as each job is complete.
Six dimensions of quality.
Our tool performs comprehensive quality analysis across six key dimensions:
Ensures the translation faithfully conveys the source meaning.
Corrects grammar, syntax, and sentence structure errors
Delivers natural, flowing sentences in the target language.
Applies your approved glossary terms consistently.
Ensures correct number formats, date formats, and measurement units.
Preserves all tags, placeholders, and special formatting.
Why choose re[words].
re[words] was born out of 2+ years of research into translation quality and language technology, developed by people who understand what translators and language teams actually need.
re[words] processes multiple files in parallel, returning corrected output in a fraction of the time manual revision requires. Shorter time-to-market. Consistent quality. No trade-off.
For teams that prefer full control, re[words] supports a BYOK (Bring Your Own Key) architecture: your API key is never stored on our servers, used exclusively during processing, then discarded. Maximum security, zero privacy compromise.
If you don’t have an API key yet, we can provide one to get you started; just reach out and we’ll help you find the right setup.
re[words] supports the XLIFF formats of the leading CAT tools on the market. No migration, no process change — just a quality layer that slots in where you already work.
Let’s talk. Our team is ready to listen and help you find the right setup for your workflow.
FAQ.
re[words] is an AI-powered post-editing tool developed by Creative Words for companies and translation teams managing high volumes of XLIFF files. It’s designed for anyone who wants to improve translation quality in a scalable, measurable way, without losing visibility over the process.
re[words] supports the main XLIFF formats used across the industry: .sdlxliff (SDL Trados), .mxliff (Phrase / Memsource), .mqxliff (memoQ), and .xlf (Okapi Framework). You can upload up to 20 files per job, with a maximum file size of 50 MB.
Yes. re[words] uses a BYOK (Bring Your Own Key) architecture: you need an active API key from OpenAI or Anthropic. The key is sent with the job request and used only during processing — it is never stored or shared. Don’t have one yet? Get in touch — we’ll help you figure out the best setup for your needs, or provide a key if needed.
You can generate an OpenAI key directly from the OpenAI developer portal, or an Anthropic key from your Claude platform settings. If you need help finding the right configuration for your workflow, get in touch — our team will get back to you within 24 hours.
Yes. For more advanced teams looking to integrate re[words] directly into their existing pipelines and systems, a dedicated API is available. If you want to automate post-editing at scale or connect re[words] to your own tools, contact us to find out more.
You can attach an Excel file (.xlsx) with your domain-specific terms. re[words] automatically injects them into every prompt, ensuring consistent terminology across all segments. The format supports up to four columns: source term, target term, context note, and entry ID.
Automatic Post-Editing analyzes an existing translation and corrects it across the selected quality dimensions. LLM Translation generates a translation from the source segments — useful for producing a first draft before human review. The two modes are designed to cover the full production cycle.
Fill in the form above or get in touch directly. Our team will get back to you within 24 hours to understand your needs and find the right configuration for your workflow.
You could open ChatGPT and paste in a translation, but that’s exactly the point: you’d be pasting, copying, formatting, repeating, for every single segment across every single file. XLIFF aPE eliminates all of that manual work and processes your files automatically, in bulk, at scale.
The prompts driving every analysis step were designed by professional linguists, who encoded translation quality expertise into structured sequences covering accuracy, fluency, style, terminology, locale, and tags. That’s not something you replicate with a generic ChatGPT session.
There’s also a data privacy consideration: consumer tools like ChatGPT may use your content for model training. XLIFF aPE connects via API, where your data stays confidential by default.
You do, and that’s by design. XLIFF aPE is not a fire-and-forget solution; human review remains a necessary step in the workflow. The TQAuditor integration lets you compare the original MT against the post-edited version side by side, making it easy to spot any issues introduced along the way.
The traditional MT workflow requires post-editing before a translator can even begin reviewing. XLIFF aPE handles that post-editing automatically, so the translator goes straight to review, working on output that has already been assessed across six quality dimensions. You’re not adding a step, you’re removing one.
Your content is processed via direct API calls to OpenAI and Anthropic — it does not pass through our servers. When you use your own API key, the data flow is entirely between your machine and the model provider. When you use our key, the same applies, with files retained for 7 days before being permanently deleted.
In both cases, the API connections used by XLIFF aPE are subject to the providers’ enterprise data policies, which do not use your content for model training by default.
It doesn’t, not out of the box, and that’s intentional. Every new client goes through an onboarding process where our AI language specialists work with you to customise the prompts driving each quality dimension. Your register, your brand voice, your terminology: all of it gets encoded into your own environment before you process a single file. This onboarding is a standard part of every subscription, not an optional add-on, because we know a generic setup isn’t good enough. The resulting prompts live in your account and are yours exclusively, of course.
XLIFF aPE is a productivity tool, not a translation service. It assists and accelerates the work of qualified professionals, but it does not replace human judgement and, as stated throughout, human review remains a mandatory step in any responsible workflow. Responsibility for the final output lies with the professional or agency delivering it.
As with any software tool, Creative Words does not accept liability for decisions made on the basis of the tool’s output. Users are responsible for ensuring that all content is reviewed and approved by a qualified professional before delivery.
