The corpus comprises speech transcripts, acoustic event identification and semantic annotation of relations holding between language and image in two TV travel series episodes. The latter comprises identification of objects, body movements, gestures and speech segments that are close in time (overlap usually) and annotation of their semantic relation according to the COSMOROE Multimedia Dialectics Framework. Object contours (numerical format) and body movement complements (agents, tools, affected objects, location of the action) are also annotated. All visual elements are tagged/labeled (verbal categorisation). Also, both transcribed speech, scene text, graphic text and tags are available in two languages, English and Greek. The annotated data can be fully navigated through the COSMOROE search engine (www.cosmoroe.eu). The resource and a repackaging of its contents for the needs of different research tasks is available with detailed readme files at: http://csri.gr/downloads/cmr