Image translator. Translate text in any image — signs, menus, screenshots, photos of documents — by running vision-OCR plus Venice translation. Returns the detected source text plus the target-language translation. Optional preserve_layout mode keeps the original spatial arrangement.