Answer
OmegaT for Professional Translators: What It Does and What It Lacks
OmegaT is the free, open-source Java CAT tool that handles TMX, XLIFF, glossaries, QA checks, and an offline /tm/mt machine-translation workflow for
OmegaT sits inside the professional translator's production stack as the free, open-source desktop CAT tool — the one that lets a freelancer run segment-level translation memory, glossaries, and QA checks without a licence server or a monthly bill. For a broader map of where it fits, see CAT Tools Compared: How Translation Memory Software Works.
Key points
- OmegaT is a free, open-source CAT tool released under GPL v3 — it costs translators exactly $0, with no per-seat fee, no annual subscription, and no licence server standing between the software and your work.
- The application runs on Windows, macOS, and Linux as a Java-based desktop program, which means a single download works across every major operating system a professional translator uses.
- Translation memory is stored as industry-standard TMX files inside the project's
/tmfolder, with dedicated subfolders (/tm/auto,/tm/mt,/tm/penalty-XX) that control how each batch of matches is weighted and surfaced in the Fuzzy Matches pane. - Fuzzy match bands follow the industry-standard tiers — 75–84%, 85–94%, 95–99%, 100% exact, and 101% in-context exact — so word-count reports OmegaT generates map directly onto common client rate cards.
- Dropping a TMX file of prior MT output into
/tm/mtgives OmegaT fully offline machine-translation matches with no API key, no cloud dependency, and no per-character billing — a particularly clean workaround on macOS, where UI-based MT plugins frequently encounter Java keystore and Gatekeeper friction. - OmegaT has been downloaded close to 900,000 times since its first release, reflecting sustained adoption by working translators and small agencies worldwide.
- Supported file formats include DOCX, ODT, HTML, XLIFF, TMX, SRT, IDML, PO, and plain text, covering the document types that appear in agency and direct-client workflows.
- OmegaT fits solo certified translation work and low-volume agency pipelines well at zero tooling cost, but lacks a built-in bilingual audit trail and server-side TM sharing out of the box — gaps that matter as team size grows.
What we offer
What is OmegaT and what does it do for a professional translator?
OmegaT is a free, open-source, Java-based desktop CAT tool that segments source files, leverages TMX translation memory and glossaries, runs inline QA checks, and exports bilingual XLIFF — giving freelancers a licence-free production surface they fully control locally. A CAT tool, as covered CAT Tool: What It Is, What It Does Inside a Translator's Workflow, is the segment-level production surface where source, TM match, term-base hit, and QA warning sit in one view. OmegaT delivers that view without a vendor, a server, or a seat fee.
What are OmegaT's key features that matter inside real translation work?
OmegaT ships six load-bearing features for professional work: segment-level TMX translation memory, glossary term bases, fuzzy matching across industry-standard bands, inline QA checks, XLIFF and TMX bilingual exchange, and a plugin API for machine-translation engines — all running locally without a licence server. Having been downloaded close to 900,000 times since its first release, OmegaT is a mature tool built by and for translators, and its feature set reflects real-world production priorities rather than marketing checklists.
| Feature | How OmegaT implements it | Where it lives on disk |
|---|---|---|
| Translation memory | TMX level 1 and 2, matched against source segments | /tm, /tm/auto, /tm/mt, /tm/penalty-XX |
| Glossary | Tab-delimited.txt and TBX, auto-highlighted in editor | /glossary |
| Fuzzy matching | 75% floor, 85%, 95%, 100%, 101% in-context bands | Fuzzy Matches pane |
| QA checks | Tags, numbers, empties, repetitions, glossary adherence | Validation window |
| Bilingual exchange | XLIFF export, TMX export | Project → Create Translated Files |
| Machine translation | Plugin API or TMX drop in /tm/mt | Machine Translation pane |
How does translation memory work in OmegaT?
Translation memory in OmegaT reads every TMX file inside the project's /tm folder, matches new source segments against stored source–target pairs, and surfaces hits in the Fuzzy Matches pane. The memory presents both 100% exact matches and fuzzy matches from any legacy translation database loaded into the folder, which directly increases translation efficiency, consistency, and quality across projects. The 75% fuzzy floor through 101% in-context bands are the same tiers the broader translation industry uses for word-count pricing, so OmegaT's output integrates cleanly into standard rate structures.
- 75–84%: low fuzzy — source and stored segment share some wording but differ enough that most of the segment needs rewriting.
- 85–94%: mid fuzzy — the match is structurally close; the translator edits rather than rewrites.
- 95–99%: high fuzzy — only minor terminology or punctuation differences separate match from source.
- 100%: exact match — source segment is identical to a stored pair; the translator verifies and confirms.
101%: in-context exact — same segment text and same surrounding segments, the highest-confidence match OmegaT surfaces.
The /tm/auto subfolder accepts TMX files whose matches OmegaT applies automatically to untranslated segments on project load. The /tm/penalty-XX subfolder accepts TMX files that are down-weighted by a configurable percentage — useful for leveraging TMs from adjacent but not identical subject areas. This folder-based architecture means a translator manages multiple TM tiers entirely through the file system, with no database administration.
How do glossaries and term bases behave in OmegaT?
OmegaT reads tab-delimited.txt and TBX glossaries from the project's /glossary folder, auto-highlights term hits in the editor as each segment opens, and lets translators add entries on the fly with Ctrl+Shift+G — no separate term-base server is required. Multiple glossary files coexist in the same folder without conflict, so a project-level TBX supplied by a client sits alongside a personal.txt word list and both remain active simultaneously. The writable glossary — the.txt file OmegaT creates automatically on first use — captures new entries immediately so terminology decisions made mid-project are available to the next segment.
What QA checks does OmegaT run before export?
OmegaT's built-in QA validation runs against the full set of translated segments and flags six classes of problem before translated files leave the project:
- Empty target segments — any segment left untranslated is listed so nothing ships as a blank.
- Tag mismatches — inline formatting tags present in the source but absent or duplicated in the target are caught before the target file is regenerated.
- Number mismatches — figures, dates, and codes that differ between source and target are flagged for review.
- Inconsistent translations of repeated source segments — the same source string translated differently across the project surfaces as a consistency warning.
- Glossary non-compliance — segments where a source term appears in the glossary but the prescribed target equivalent is absent are highlighted for correction.
- Optional LanguageTool grammar and style checks — when the LanguageTool plugin is active, grammar, style, and punctuation rules across more than 30 languages run automatically against target text.
How do you set up OmegaT and start a translation project?
Setting up OmegaT takes four steps: install Java 11 or later plus the OmegaT package for your OS, create a new project with a language pair and the default folder layout, drop source files into /source, and press Ctrl+T to generate translated target files — the whole flow runs offline.
flowchart TD A[Install Java 11+ and OmegaT] --> B[Create new project] B --> C[Pick source and target language] C --> D[Drop files into /source] D --> E[Translate segments] E --> F[Ctrl+T creates /target files]
What does the standard OmegaT project folder structure look like?
A standard OmegaT project contains seven folders plus one project file, each with a defined role:
| Path | Role |
|---|---|
| /source | Input files to translate |
| /target | Exported translated files |
| /tm | Human translation memories (TMX) |
| /tm/mt | Machine-translation TMX, treated as MT matches |
| /tm/auto | TMX pre-applied to matching segments on open |
| /tm/penalty-XX | Discounted TMs where XX is a fixed penalty percentage |
| /glossary | Term bases (.txt tab-delimited or.tbx) |
| omegat.project | XML config file for the project |
How is OmegaT installation different on Mac, Windows, and Linux?
OmegaT installs identically across Mac, Windows, and Linux once a compatible JRE is present. The platform differences are narrow:
- Windows: run the installer, accept defaults.
- macOS: right-click the app and choose Open on first launch to bypass Gatekeeper. The bundled-JRE package is the lower-friction path.
- Linux: unpack the tarball and run the shell script, or install from distribution repositories where available.
How do you use machine translation with OmegaT, including offline on Mac?
There are two ways to feed MT into OmegaT: the plugin UI connecting to Google, DeepL, Microsoft, Yandex, or Apertium over API, or the offline /tm/mt workflow — drop a TMX of prior MT output into the project's /tm/mt subfolder and OmegaT surfaces it in the Fuzzy Matches pane as MT matches, no API key or internet needed.
flowchart TD
A[Need MT in OmegaT?] --> B{Online and plugin working?}
B -- Yes --> C[Options -> Preferences -> MT, paste API key]
B -- No --> D[Pre-generate MT output as TMX]
D --> E[Drop TMX into /tm/mt]
E --> F[Fuzzy Matches pane shows MT matches offline]
C --> G[Translate segments]
F --> G
How do you connect Google, DeepL, or Microsoft MT in OmegaT's UI?
To connect a cloud MT engine, open Options → Preferences → Machine Translation, enable the engine, paste the API key, and tick 'Automatically fetch translations' — OmegaT then calls the engine per segment and shows results in the Machine Translation pane. Each engine needs its own account and billing arrangement, which stays outside OmegaT.
Can you use MT offline in OmegaT without a live API connection?
You can use MT offline in OmegaT by pre-generating a TMX file from the MT output of your choice and placing it inside <project>/tm/mt — OmegaT treats the file as machine-translation matches and surfaces them in the Fuzzy Matches pane without any live API call. This is the fully offline MT integration no competitor documents.
Where do you put TMX files for MT in OmegaT?
TMX placement decides the match's semantic class — the folder, not the file contents, labels the match.
| Folder | Semantic class | Where it appears |
|---|---|---|
| /tm | Human translation memory | Fuzzy Matches pane, full leverage |
| /tm/mt | Machine translation | Machine Translation pane, no TM leverage |
| /tm/auto | Pre-applied matches | Written straight into segments on project open |
| /tm/penalty-XX | Discounted human TM | Fuzzy Matches pane, scores reduced by XX% |
Why is this TMX workflow particularly useful on Mac?
On macOS the UI-based MT plugins hit Java keystore trust issues and Gatekeeper friction more than on Windows. The TMX-in-/tm/mt route bypasses the plugin layer entirely — the Mac translator pre-generates MT output elsewhere, saves it as TMX, drops it in /tm/mt, and works fully offline. No keystore import, no plugin signature prompt, no API key stored in a notarised binary.
How do you separate MT matches from human TM matches in OmegaT?
OmegaT separates MT matches from human TM matches by folder: /tm/mt entries display in the Machine Translation pane and never dilute human-TM leverage economics. A reviewer scoring segments can tell at a glance which came from an engine and which from a prior human translation. This matters because MT matches do not qualify for the 75% fuzzy floor pricing — treating them as TM corrupts the rate card.
How does OmegaT compare to Trados, memoQ, Phrase, and Wordfast for production work?
OmegaT trades vendor polish for freedom. For the broader landscape, see CAT Tools for Translation: What They Do and Which Ones Translators Actually Use.
| Tool | Licence cost (US, 2025) | Deployment | Server TM | Certified-lane fit |
|---|---|---|---|---|
| OmegaT | $0, GPL v3 | Desktop | No native | Low-stakes solo only |
| SDL Trados Studio | $400–$900 perpetual freelance | Desktop (Trados Team cloud) | Yes | Primary choice |
| memoQ | $620–$900 perpetual | Desktop + optional server | Yes | Primary choice |
| Phrase | $27–$80 per user per month | Cloud / browser | Yes | Weak for offline certified work |
| Wordfast | Lower seat cost than Trados | Desktop | Limited | Freelance, light certified |
What does OmegaT do that the paid CAT tools do not?
OmegaT keeps three advantages the paid tools cannot match:
- Zero licence cost, now and forever, under GPL v3.
- Full local control of every file — nothing leaves the workstation unless the translator ships it.
- A transparent plain-folder project structure that any text tool can inspect without a proprietary database.
What are OmegaT's honest gaps versus paid CAT tools?
OmegaT lacks four things the paid tools provide:
- Server-side TM sharing across translators in real time.
- A vendor support SLA — no ticket queue, no account manager.
- Deeper QA heuristics tuned by LSP consortia and shipped as default profiles.
- Native SDLXLIFF and MQXLIFF round-trip — agency jobs shipped as SDLXLIFF need conversion first.
Does OmegaT fit certified translation production — and where does it stop fitting?
OmegaT fits low-stakes certified translation work where the translator owns the whole lane: it handles segment edits, TM, glossaries, and QA cleanly, but cannot produce a vendor-signed bilingual audit trail without manual rigging. USCIS filings at $0.10–$0.25 per source word still need the certificate of accuracy attached as a document step, not a CAT feature. The optional notarization line adds $10–$50 on top.
What does the LQA gate expect that OmegaT does not deliver natively?
The LQA gate samples translated output at $40–$90 per 1,000 words against an MQM-style scorecard aligned to ISO 17100. The scorecard covers accuracy, fluency, terminology, locale conventions, and style. OmegaT's built-in QA covers tags, numbers, and glossary adherence — the MQM categorisation and sampling is a separate reviewer workflow outside the tool.
When is OmegaT the right tool for a USCIS or court translation?
OmegaT is the right tool for a USCIS or court translation when three conditions hold at once:
- The translator is working solo, with no second-pair-of-eyes sign-off inside the CAT.
- The source is a single-file personal record — birth certificate, academic transcript, affidavit.
- The certificate of accuracy is produced as a separate signed PDF attached to every page, priced into the $25–$150 per-page rate.
What are OmegaT's limitations translators run into after a few projects?
After a few projects, OmegaT translators encounter five recurring limits that become more visible as workload and team size grow. None of these limitations is fatal for solo or small-agency work, but each represents a genuine gap relative to server-based CAT platforms.
- No live concordance across remote corpora. Concordance search in OmegaT is local-TM only — it scans the TMX files currently loaded in the project. Translators working across dozens of past projects cannot search a centralised, always-current corpus without first consolidating TMX files manually.
- No shared term-base locking in team work. Because glossary files are plain.txt or TBX files on disk, two translators writing to the same glossary simultaneously can overwrite each other's entries. A shared-drive or version-control workaround is possible but is not built in.
- Java startup lag on large projects. OmegaT is a Java-based desktop application, and projects that load hundreds of megabytes of TMX data across multiple subfolders can take meaningful time to initialise, particularly on machines with constrained RAM.
- Limited DOCX round-trip fidelity on complex layouts. Segments translate correctly, but source files that contain nested tables, anchored shapes, or heavy custom styles do not regenerate with pixel-perfect formatting. Simple to moderately complex DOCX files round-trip reliably; heavily formatted documents need post-processing in the word processor.
- No built-in connector to agency TMS platforms. Handoff to project-management platforms is manual XLIFF export and import. There is no native plug-and-play integration that pushes translated files back to an agency server automatically, which adds steps to every client delivery cycle.
Translators who need server-side TM sharing, real-time collaborative glossary management, or automated TMS round-trips will find that OmegaT's zero-cost model trades those conveniences for full local control over data and workflow. Understanding where those trade-offs land is the key to deciding whether OmegaT fits a given project type or client relationship.
How we use OmegaT in our own production
We treat OmegaT as the right tool for a specific slot: solo, offline, low-stakes certified work where the translator and the certifier are the same person. The /tm/mt folder is where we pre-stage machine-translation output as TMX before a project opens, so the editor never calls a cloud engine mid-segment.
We decline to use OmegaT as the primary surface for agency SDLXLIFF jobs that ship with a vendor-signed audit expectation. The round-trip through Okapi Rainbow works, but every conversion step is a place a signature can drop, and signatures are the thing an LQA reviewer actually scores. The tool that owns the SDLXLIFF format should own the file.
We also decline to call OmegaT's QA pane "LQA ready." It flags tags, numbers, and glossary gaps. An MQM scorecard categorises errors by severity against ISO 17100 — that is a reviewer workflow, not a QA button, and claiming otherwise would mislead a buyer who is paying $40–$90 per 1,000 words for the gate.
The questions below cover the four workflow questions translators ask most about OmegaT — Mac MT setup, offline MT, TMX placement, and plugin-free MT configuration — plus the licence, format, and certified-lane questions that follow from them.
Frequently asked questions
How do I use machine translation with OmegaT on Mac?
On Mac, OmegaT's machine translation setup follows the same steps as on Windows or Linux — you configure your MT engine under Preferences > Machine Translation. For cloud-based engines such as DeepL or Google Translate, you enter your API key in that panel and enable the plugin; OmegaT then displays MT suggestions in the Machine Translation pane alongside your fuzzy matches and glossary hits. Because OmegaT is Java-based and cross-platform, there are no Mac-specific menus or file paths to worry about — behavior is identical across operating systems.
Can I use MT offline in OmegaT?
Yes, OmegaT supports offline machine translation through locally hosted engines such as LibreTranslate or Argos Translate, as well as self-hosted instances of models like OPUS-MT. You point OmegaT's MT plugin to a local API endpoint (typically localhost on a port you configure) instead of a cloud URL, so no data leaves your machine. This is especially relevant for certified translation workflows where source documents may contain confidential personal, legal, or medical content that cannot be sent to third-party servers. Setup requires running the local engine as a service before launching OmegaT.
How do I set up MT suggestions in OmegaT without going through the UI?
MT engine settings in OmegaT are stored in the omegat.prefs XML file inside your OmegaT user configuration directory, so you can edit them directly in a text editor without touching the UI. The configuration folder is found at ~/.omegat/ on Mac/Linux or %APPDATA%\OmegaT\ on Windows. You can add or update the relevant plugin keys — such as your API key and engine URL — directly in that file, then restart OmegaT for changes to take effect. This approach is useful for scripted deployments in agency environments where multiple translators share a standard configuration.
Related
- Pharma Localization: What Translators and Agencies Actually Deliver
- Phrase AI: What It Does for Translators and Where It Fits the Certified-Translation Stack
- Phrase CAT: What the Editor Does Inside a Translator's Workflow
- Product Localization: How Translators and Agencies Deliver It
- Pseudo Localization: What Translators Need to Know About i18n Test Builds
- React i18n: How Translators Fit Into a React Localization Pipeline