We propose a project in affective speech synthesis in which we take two orthogonal approaches to synthesis of affect. In the first, data driven approach, we propose to learn affect using deep learning. In the second approach, we propose to use parametric approaches to prosody and formant shifting. We also plan to cover the main Swiss languages; these include German, in which much previous work on affect has been done, and Italian, which is intuitively more emotive, but has previously been overlooked in this and related subjects.