Machine translation (MT) is the automated process of converting text from a source language into a target language using AI models, and it plays a central role in multilingual course production by enabling localization at a speed and cost that human-only translation cannot match. Modern MT systems — led by neural machine translation (NMT) engines such as DeepL, Google Translate, and Azure Translator — produce output quality that is highly usable for many language pairs, especially in general-register prose, though quality degrades with highly specialized domain terminology, idiomatic expressions, and low-resource languages. In e-learning workflows, MT is applied to ASR-generated transcripts to produce translated subtitle files, to course scripts before TTS re-narration in a new language, and to on-screen text overlays. The standard professional workflow is machine translation followed by post-editing (MTPE) by a human translator, which is faster and cheaper than translation from scratch. A key gotcha for learning video is that translated subtitles must remain synchronized with the original audio timing; MT does not preserve timing data, so subtitle timing must be adjusted after translation. Maintaining a translation memory — a database of previously approved segment translations — speeds up MTPE on future course updates and enforces consistent terminology across a curriculum. Cultural adaptation — not just linguistic translation — matters for learning effectiveness: examples, humor, and regulatory references may need to change even when the words translate accurately.
