In this paper we present a method and a tool for modelling a teletraining session in heterogenous, distributed open environments. We propose a mathematical notion for the training process. Therefore we divide a whole training session into presentation units, define some relations on these units and develop a controller for running the session. Units consist of multimedia objects, such as text, graphics, video and audio, which have to be displayed with time and space synchronisation and coordination. In a related project at the Department of Software Engineering at the Upper Austria Polytechnic University a tool for modelling and running such teletraining sessions was developed. It is based on the herein defined formalism and compiles well defined unit models into JAVAT/sup TM/ code, which can be executed by usual WEB-Browsers.