This paper describes the prototype of a spoken conversational agent embedded within a simulation environment. This prototype accepts utterances from a user enabling him or her to navigate into a three-dimensional image of brain. The paper first describes what we can expect from such an interface in the communication quality between a user and represented in virtual worlds entities – artifacts –. Images have been obtained using magnetic resonance techniques. To enable a fast animation of images, they are reduced to surface elements. The paper describes how the images were processed and reconstructed. Then it describes the prototype’s architecture which includes a speech recognition device together with a speech synthesizer. The system consists in a chart parser for spoken words; a semantic analyzer; a reference resolution system; a geometric reasoner, and a dialogue manager. The prototype has been implemented and has been demonstrated.