This paper introduces an innovative chatbot designed to comprehend and interact with users about visual content, particularly images. The chatbot seamlessly merges natural language processing (NLP) and computer vision, enabling it to process user inquiries, analyse images, and generate insightful responses. The underlying model employs a neural encoder–decoder architecture with a late fusion encoder, which effectively integrates image features and textual representations. The chatbot utilizes two distinct decoders: a generative decoder for crafting natural language responses, and a discriminative decoder for evaluating response relevance. Incorporating object detection techniques, notably Mask R-CNN, empowers the chatbot to pinpoint and identify specific objects within images, leading to more focussed and informative answers. Rigorous training and evaluation on the COCO dataset showcase the chatbot’s proficiency in understanding and communicating visual content. This research paves the way for the development of more interactive and intelligent chatbots capable of engaging in meaningful conversations about images, with potential applications spanning diverse fields.