XLIFF 1.2 · XLIFF 2.0 · AI translation

Translate XLIFF files online. Generic XLIFF, real tag handling.

Drop any XLIFF — exported from Trados, memoQ, Phrase, custom systems, or your own pipeline. Fily handles <source>, <target>, <g>, <x>, <ph> and 2.0's <segment> / <unit> model without breaking on import.

0XLIFF jobs0words processed

What is XLIFF (Generic)?

XLIFF (XML Localization Interchange File Format) is the OASIS-standard format for exchanging bilingual content between translation tools. Two major versions are in use: XLIFF 1.2 (since 2008, used by Trados and most LSP-grade interchange) and XLIFF 2.0 (since 2014, restructured around <file> → <unit> → <segment>).

Generic XLIFF (vs vendor-extended like SDLXLIFF or MQXLIFF) is what you exchange when one side of the workflow doesn't run a specific CAT tool — typically an LSP receiving content from a custom CMS, a translator working without Trados/memoQ, or an interchange between platforms.

Why XLIFF (Generic) is tricky for AI translation

  • Version detection: 1.2 and 2.0 have incompatible structures. Tools that don't auto-detect produce empty target segments or invalid output.
  • State attributes: XLIFF 1.2 uses state="translated|reviewed|signed-off"; 2.0 uses different state machines. Resetting these on round-trip breaks downstream workflow.
  • Inline tags: both versions allow <g>, <x>, <bpt>/<ept> (1.2) or <pc>, <ph>, <sc>/<ec> (2.0). Mixing them or losing IDs invalidates the file.
  • Namespace soup: custom XLIFF often pulls in ITS, XLF extensions, or vendor namespaces. A naive parser strips them and the receiving system rejects the file.
  • Plurals and segmentation: XLIFF 2.0 supports complex segmentation with multiple <segment> per <unit>. AI tools that flatten this lose plural-form information.
  • Notes and comments: <note> elements at trans-unit level often carry translator instructions ("do not translate", "max 30 chars"). Stripping them loses critical context.

How Fily handles XLIFF (Generic)

  • Auto-detect version: XLIFF 1.2 vs 2.0 detected from root element + namespace declaration. Routing to the correct internal pipeline.
  • Namespace-preserving parser: custom namespaces and ITS extensions pass through unchanged.
  • State-attribute pass-through: state and state-qualifier from the source are preserved unless the segment is actually retranslated.
  • Inline-tag codec: <g>, <x>, <ph> (1.2) and <pc>, <sc>, <ec> (2.0) extracted with full attribute sets, segments translated as text with placeholders, tags reinserted.
  • Note preservation: <note> elements at file, group, and unit level are read into the translation prompt as instructions (so "do not translate" is honored) and written back unchanged.
  • Segmentation respect: multi-segment units in XLIFF 2.0 are translated segment-by-segment, not flattened to unit-level.

Pipeline: xml_qa_12step_v2@1.0.0

The XLIFF (Generic) workflow with Fily

1

Upload

Drop your .xliff (single or batch ZIP). Optional: glossary, TM, style guide.

2

Process

Fily runs the XLIFF (Generic) pipeline + 12 QA steps. Typical job: 10–20 minutes.

3

Download

Same format, ready to deliver. QA report HTML attached.

LSPs using generic XLIFF typically work between systems: a client's custom CMS exports XLIFF, the LSP translates, the CMS re-imports. Fily fits the LSP side of that loop — translators get pre-translated files with notes honored, reviewers approve.

Beyond the standard pipeline

What we've built around XLIFF (Generic)

Edge cases clients brought us for this format — and the pipelines we shipped to solve them.

Golden TM · auto-build

A translation memory that builds itself

Clients wanted TM leverage but didn't want to maintain a TM — exporting, cleaning and re-importing TMX after every project is a job nobody has time for.

Every confirmed segment is promoted into the organization's Golden TM automatically, scoped per client and language pair, and reused on every future job. Reviewer edits in the dashboard feed straight back in. Exact matches then cost $0.

TMX Aligner

Turn a finished translation into reusable TM

A new client had years of past translations but no translation memory — just source documents and their finished translations sitting in folders, in no bilingual format.

The aligner takes a source file and its existing translation — two plain files — and returns a ready-to-use TMX plus an extracted glossary. Day-one leverage from an archive that was doing nothing.

Frequently asked about XLIFF (Generic)

Ready to translate a XLIFF (Generic) file?

No card. No setup. Upload one file and see the output.