We describe the tasks that may be performed by the fission module in a system that generates multimodal output. Broadly speaking, these tasks fall into three categories: content selection and structuring, modality selection and output coordination. We summarise the requirements of each of these tasks, and the approaches taken to them by a number of existing multimodal presentation systems. We also give an overview of two of the specific output techniques that will be used in COMIC: speech synthesis and embodied conversational agents. We describe several languages that have been used used to specify the content of multimodal presentations or that have been proposed as standards for this purpose. Finally, we analyse the existing work in the context of COMIC and make some recommendations.