Translate Phrase TXLF files online. Tags intact. Review-state respected.
Drop your .txlf exported from Phrase TMS and get it back fully translated, with Phrase's inline tags preserved and TM matches passed through untouched.
What is Phrase TXLF?
TXLF is the bilingual export format of Phrase TMS (formerly Memsource), one of the most widely used continuous localization platforms. It's an XLIFF 1.2 variant with Phrase-specific extensions: inline tags as <bpt> / <ept> / <ph> pairs, segment confirmation states (unconfirmed, translated, confirmed, reviewed), match metadata from Phrase's TM, and machine-translation pre-fills with scoring.
Phrase TMS is optimized for continuous localization — software strings, marketing content, documentation that updates frequently. TXLF exports are typically used when external translators or LSPs work outside the Phrase web editor, then return files for re-import into the Phrase project.
Why Phrase TXLF is tricky for AI translation
- Phrase tag IDs use compact integers (<ph id="1"/>) that must remain unique within each segment and preserved across source-target.
- Pre-filled MT suggestions appear as already-populated target segments with match scores. AI tools that "improve" these segments destroy the original MT provenance and break Phrase's quality analytics.
- Confirmation state propagation: a segment marked confirmed in source should not be reset to unconfirmed on return, unless the translation actually changed.
- TBX terminology: Phrase projects ship with TBX-format term bases that need to be honored — terms with forbidden or preferred flags must be enforced in the target.
- Aggressive review-mode promotion: some AI pipelines promote any review-status segment to full re-translation, wasting tokens and breaking client-approved content.
- In-context matches (101%): Phrase distinguishes 100% (exact) from 101% (in-context exact). Both must be passed through, but the in-context status must survive.
How Fily handles Phrase TXLF
- Phrase-aware tag codec: <ph>, <bpt>, <ept> are extracted with their numeric IDs, segments translated as text with placeholders, tags reinserted with original IDs.
- Opt-in aggressive review: by default, review-state segments are passed through untouched. The aggressive promotion behavior is opt-in — disabled for every standard job.
- Match-score pass-through: 100% and 101% segments are never sent to the AI. 101% in-context status is preserved.
- TBX terminology enforcement: term bases attached as TBX are parsed and applied via Fily's glossary enforcement step. Forbidden terms are replaced.
- Confirmation state logic: confirmed segments are not re-translated unless explicitly requested. New translations are marked translated (not confirmed) so reviewers see what changed.
- 12-step QA: glossary, forbidden, tag repair, consistency, no-source-bleed, and 7 more passes.
Pipeline: phrase_docx_v2@1.0.0 (TXLF native handling)
The Phrase TXLF workflow with Fily
Upload
Drop your .txlf (single or batch ZIP). Optional: glossary, TM, style guide.
Process
Fily runs the Phrase TXLF pipeline + 12 QA steps. Typical job: 10–20 minutes.
Download
Same format, ready to deliver. QA report HTML attached.
LSPs working with Phrase clients typically receive TXLF batches when the client's internal team can't keep up with volume. Running them through Fily before assigning to reviewers cuts the post-edit time by 60–75% — reviewers approve segments rather than translate them from scratch.
Frequently asked about Phrase TXLF
Other formats Fily translates
Ready to translate a Phrase TXLF file?
No card. No setup. Upload one file and see the output.