We create a retrieval engine system using content based image retrieval techniques and fusing them with text annotations captioned from the web. This captions will be automatically selected starting with a vocabulary.