We enhance visible-infrared person re-identification by leveraging attribute arrays extracted using a large vision-language model.