Machine translation post editing: what MTPE is and how the work is done

What post-editing is

In MTPE, the machine produces a complete first draft and a translator edits it into the finished text. The translator reads the source and the draft side by side, keeps what is right, fixes what is wrong and rewrites what reads like a machine. The skill is deciding fast which sentences to leave alone.

Post-editing is now a common service. In the survey of 777 freelancers by the SFT, the French translators' association,, 59% offered post-editing, mostly to agencies rather than direct clients[2].

Light versus full post-editing

The industry splits post-editing into two levels, and the client should name one before the job starts.

Full post-editing

The output should be equivalent to human translation. ISO 18587 sets its requirements, and the post-editor works on accuracy, fluency, terminology and style[1]

Light post-editing

The goal is a text that conveys the meaning of the source accurately. Minor style and punctuation issues are left alone[1]

Who sets the level

Client and translator, per job, from the text's purpose and target audience[1]

Light post-editing suits internal reading, such as a supplier's letter or a manual a team only needs to understand. Full post-editing suits anything published, filed or signed. A certified translation for an authority is full post-editing at the very least, because the translator signs that it is complete and accurate.

How MTPE is priced

Most post-editing is still billed per word. The SFT survey found word rates are the dominant pricing method for post-editing, even though respondents rated hourly pricing as the most satisfying for it[2]. The median agency rate was €0.05 per source word, against €0.09 for translation[2]. Freelancers who post-edit for direct clients charged around €50 an hour, about the same as revision[2].

A per word rate assumes the machine draft is good. When the source is a poor scan, the draft inherits every OCR error, and the post-editor ends up retranslating. That is the case for an hourly rate, or for fixing the source text before the machine ever sees it.

For rates in other markets and units, see the translation rates guide.

A post-editing workflow that starts from a scan

Clients often send a scanned contract, certificate or report and ask for MTPE pricing. LoreAI covers both the OCR and the machine draft in one step.

1. Upload the scan to LoreAI. PDF, JPG and PNG work. LoreAI reads printed text, handwriting and stamps with OCR.
2. Choose the target language. The source language is detected automatically. You get a translated Word file with headings, lists and tables in the order of the original, and names, dates and numbers as written.
3. Post-edit in Word or in your CAT tool. Edit the Word file directly for a short document. For a longer job, also run OCR Only on the scan, import the source file into your CAT tool and use LoreAI's translation as reference.
4. Deliver at the agreed level. Light or full, as named in the job.

On paid plans, the CAT-ready Word option builds the file for clean segments in Trados, memoQ, Phrase or Smartcat. What a CAT tool is explains the import side. Your first document is free, up to 3 pages.

When post-editing is the wrong service

Some texts are cheaper to translate than to post-edit. Marketing copy, wordplay and anything where tone matters most often needs a rewrite, not an edit. Ask for a sample page of the machine draft before you agree to an MTPE rate, and price the job by what the sample shows.

Translators also disagree about the work itself. In the SFT survey, income satisfaction was lower for post-editing than for translation and revision[2]. Knowing that, many quote post-editing only for text types where the draft is consistently strong.

Questions people ask