With the rising popularity of and legal requirements for media accessibility, a need has arisen to train professional audio describers and, to this end, to educate audio description trainers. This paper aims to serve as guidance for trainers who wish to design and conduct their own audio description course. It first identifies shortcomings in the current audio description training literature. We then specify theoretical foundations for course design, including situated and contextualised learning models. Building on the experience from the ADLAB PRO project, we then present a procedure for designing an audio description course based on learning outcomes and such elements as learning through acquisition, inquiry, discussion, practice and collaboration. This typology of course elements serves as the analytical framework for examples of good practices described in the article. It is the first study to date that shows how audio description training can meet the demands of modern learning models and the requirements of an emerging profession. It can be used as a reference point by all those creating curricula and by trainers willing to enrich their teaching approaches.