Translate from the transcript (not the VSE) for most accurate results
You can translate to 100 languages from either the transcript (the screen which has paragraphed text) or the visual subtitle editor (where your text is split as it will appear as subtitles/captions on-screen).
BUT for increased accuracy we recommend you translate from the transcript, not the visual subtitle editor.
This is because translation is done a chunk at a time. With paragraphed text, the chunk is bigger giving the A.I. more context for accuracy when translating. When it is in the visual subtitle editor, the subtitle card has relatively fewer words giving the translation A.I. less context to translate from; accuracy will be lower.
===
To learn more of the difference between the transcript editor vs the visual subtitle editor, click here.