Light vs. Full Post-Editing

Light post-editing (MTPE-L): the editor corrects only clear errors — mistranslations, broken text, severe fluency failures — without improving style or naturalness beyond minimum acceptability. Appropriate for internal documentation, rough drafts, and content with very low visibility. Full post-editing (MTPE-F): the editor brings the MT output to the same quality as human translation — correcting all errors, improving fluency, aligning with style guide. The output is indistinguishable from human translation; the process is faster and cheaper than human translation from scratch.

Post-Editing in Game Localization

Post-editing is appropriate for specific game content types: system messages, UI strings with standardized patterns, patch notes, and high-volume documentation. Post-editing is not appropriate for: narrative dialog, marketing copy, creative content, or any text where authorial voice and naturalness are primary goals. The application of MTPE should be documented per content type — mixing MTPE and human translation without disclosure creates inconsistent quality expectations.

The Post-Editor’s Role

Post-editing requires different skills from translation. A good post-editor must: rapidly identify MT errors (including subtle mistranslations that read fluently), distinguish errors from acceptable MT output (over-editing wastes time), and apply consistent style guide decisions across high volume. Experienced post-editors work faster than translators on MT-quality content, but slower on poorly performing MT content. Language pair and content type significantly affect MT quality and therefore post-editing effort.

When MT + Post-Editing Is Not Faster

A common misconception: MTPE is always faster than human translation. This is only true when MT quality is high. For language pairs with poor MT performance (Japanese, Korean, Arabic into European languages; literary and creative content in any direction), post-editing MT output can take as long as — or longer than — translating from scratch, because the editor must correct more errors than they would produce in a clean translation. We assess MT quality before committing to MTPE workflows.

SandVox and Post-Editing

SandVox applies MTPE selectively — only for content types where MT quality is consistently high enough that post-editing genuinely saves time and cost. We document exactly which content types receive MTPE vs. full human translation on every project.

Related terms: Mtpe · Translation Memory · Game Localization · Linguistic Testing

Frequently Asked Questions