XML and DITA translation that respects your markup

Read about DeepL's quality scores, including an 80–90% cost reduction for organizations adopting AI-native platforms, per Nucleus Research.

XML files have a distinct structure, and that affects how our Language AI tool translates them. Keep these factors in mind before uploading your files.

• Comments may be treated as translatable text: Content inside "!-- comments --" can be picked up for translation, with the comment delimiters escaped in the output. Strip comments before uploading if downstream tooling depends on them being intact.
• Move non-translatable content out of CDATA blocks first: Because CDATA content is translated by default, you should move out code, regex, or other non-natural-language content inside a CDATA block before upload
• Files using ITS 2.0 external rules currently fail: Files referencing external rules via its:rules/@xlink:href return an HTTP 500 error. Inline the rules or remove the reference before uploading.
• Malformed XML returns a server error, not a validation error: Invalid XML currently returns HTTP 500 rather than a clearer 4xx error. Validate your file with a standard XML parser before uploading to catch issues early.

DITA topic files (.dita) get the same tag-aware translation as XML. These are a few format-specific behaviors that technical writing and documentation teams should be aware of.

• Only topic files are supported, not maps: .dita topic files translate directly; .ditamap files are rejected and need to be handled topic by topic
• translate="no" is fully honored: This applies on both inline and block-level elements for precise control over what DeepL AI translates
• Content references aren't resolved: conref and conkeyref references aren't followed automatically. You need to translate each referenced source topic independently.
• Code samples may lose line breaks: Newlines inside "codeblock", "pre", "msgblock", and "screen" elements may collapse to single spaces. Move code samples out of the topic before translation if line breaks matter.
• Densely marked-up topics may fail: Topics combining multiple inline elements ("filepath", "cmdname", "option", nested "indexterm") in close proximity may return an internal error. Simplify or split the topic if this happens

Go to the web translator and follow these steps to translate your file:

• Log in to your DeepL Pro or DeepL API account
• Select the "Translate files" tab
• Upload or drag-and-drop your XML or DITA file
• Set your source language, or let DeepL automatically detect it
• Select your target language
• Click Translate, and wait for completion
• Download your translated XML or DITA file

XML and DITA exist because technical documentation needs more structure than plain text. DeepL translates that structure without disturbing it.

• Structured-content and technical documentation teams: Translate product manuals, API references, or data feeds built on XML, with markup and metadata preserved
• Technical writers migrating from FrameMaker to DITA: Translate individual DITA topics with the same element-level control your existing authoring workflow relies on
• Medical device and regulated-industry documentation teams: Translate compliance-critical technical documentation with the structural precision regulatory review demands

Your technical documentation rarely lives in one file. We crafted DeepL Pro to handle the whole set.

• Translate PDF, PPT, Word, Excel, HTML, SRT, TXT, XLIFF, JSON, XML, DITA, MIF and IDML files
• Translate multiple files and into multiple target languages at once
• Use a glossary so technical terms stay consistent across topics and files
• Activate formal or informal tones, fine-tuning your translation to different audiences
• Benefit from maximum data security

Need to translate PDFs, images, and other file types? DeepL’s Language AI adapts visual content to match, so your translations stay consistent and correctly formatted no matter which file you're working with.