Lionrapid

Join the waitlist

Product · CAT Editor

An editor that carries the whole toolchain

Translation memory, termbase, machine translation, and quality checks live where the work happens: on the segment, while you type. No plugin shopping, no tab juggling.

In-editor translation memory and QA loopA segment receives translation memory matches and machine translation suggestions, the translator edits it, live QA detectors check the result and offer auto-fixes, and the confirmed segment is written back to the translation memory.SegmentSuggestTM match · MTEdittranslatorLive QAtags · numbers · termsConfirmauto-fix appliedconfirmed segments update the translation memory

Translation memory

Never translate the same sentence twice

Attach translation memories to any project and the editor surfaces exact and fuzzy matches on every segment. Confirmed segments flow back into the TM, so your leverage grows with every job.

TMs are portable: import and export TMX, and reuse memories across projects and organizations.

app.lionrapid.com
The LionRapid CAT editor: side-by-side segments with per-segment quality-estimation scores and QA warnings

Termbase

Terminology that understands morphology

Glossary terms are highlighted in the source as you work, including inflected forms. Term matching is lemma-aware in around 35 languages, so heavily inflected words don't slip past your terminology.

Spellcheck runs through Hunspell in 13 European languages, plus Finnish via Voikko. Finnish goes further still: a real grammar checker, and a glossary autofix that inflects the mandated term to match the case and number of the sentence it lands in.

Import TBX termbases, build project glossaries, or add terms without leaving the editor.

  • TBX import
  • Lemma-aware matching, ~35 languages
  • Hunspell spellcheck, 13 European languages
  • Voikko spellcheck & grammar (Finnish)
  • System termbase included
  • Add terms from the editor

Machine translation

Seven engine modes, one guardrail

MT is a per-project choice with seven engine modes: Google Translate, DeepL, self-hosted large language models, a default mode with automatic fallback between them, plus copy-source and none for the content that should never touch an engine. Exact TM matches are reused before MT is even asked, and segments above a TM-match threshold can skip MT entirely.

Whatever the engine, entity protection recognizes 13 entity kinds: nine deterministically, and four (people, organizations, locations, product names) by an LLM on our own servers. It masks the nine that must survive verbatim: emails, URLs, IBANs, phone numbers, product codes, acronyms, people, organizations, and product names. Dates, money, and measures are deliberately left to localize, and place names to translate; a QA check audits that the numbers still agree. Masked entities are restored verbatim after the engine and audited by QA; the pipeline fails soft if the LLM layer is unavailable.

Bad MT can be flagged per segment, and an LLM repair chat helps fix a stubborn sentence without leaving the row.

Entity protection around machine translationNamed entities such as brand names, product codes, IBANs and URLs are masked before machine translation, restored verbatim in the output, and audited by QA checks that flag any dropped or leaked entity.Sourcebrands · codes · IBANsMaskentities lockedMTnever sees entitiesRestoreverbatimQA auditdropped? leaked?what must not be translated never reaches the engine

Formatting & safety

Bold stays bold, versions stay recoverable

Inline formatting (bold, italic, underline) is mapped through translation instead of being flattened. Snapshots and versioning keep every state of the document recoverable.

  • Inline formatting preserved
  • Editor snapshots
  • Versioning
  • Segment states tracked

Try the editor on your own file

Upload a document, attach a TM, and see the suggestions appear.

Join the waitlistTalk to us