TMX format (.tmx)

TMX (Translation Memory eXchange) is the XML interchange format for content memories. The expansion of the acronym is the LISA standard's own name, not our vocabulary, so it keeps its original wording — a format we do not own is not ours to rename. A TMX document holds a header declaring the source language (srclang) and a body of <tu> translation units; each unit carries one <tuv> translation-unit variant per language, with optional unit- and header-level <note> and <prop> metadata.

The reader emits one translatable block per <tu>: the source segment matches the header srclang and a target segment is produced for each other-language <tuv>. The header attributes and any header-level note or property are surfaced as a single non-translatable data part ahead of the blocks, while unit-level notes and properties attach to the block. This makes TMX usable both as an input for review or alignment and as an output for exporting a memory. It mirrors Okapi's TMX filter.

IDtmx

SourceBuilt-in

Extensions.tmx

MIME Typesapplication/x-tmx+xml

CapabilitiesRead + Write

Parameters

Parameter	Type	Default	Description
codeFinderRules	array		Regex patterns that match inline codes within translatable text
escapeGT	boolean	false	Escape > as > in output XML
exitOnInvalid	boolean	false	Stop processing when encountering invalid TUs. When false, invalid TUs are skipped.
processAllTargets	boolean	true	Read all target language TUVs from each TU. When false, only the first target is read.
useCodeFinder	boolean	false	Enable regex-based inline code detection in translatable text

Configure it live

Configuration

Loading form…

Configuration output

vs defaultsno changes

apiVersion: neokapi/v1
kind: TmxFormatConfig
metadata:
  name: tmx
spec:
  # (default configuration)

Examples

Read only the first target

Reduce a multilingual memory to a single target language.

processAllTargets: false

Fail fast on invalid units

Abort processing when a malformed translation unit is found.

exitOnInvalid: true

Processing notes

• The source segment matches the header srclang; one target segment is produced per other-language <tuv>.

• Header attributes and header-level notes and properties are emitted as a single non-translatable data part ahead of the blocks; unit-level notes and properties attach to the block.

Reference documentation

← Back to the Format Reference