This presentation highlights our recent research conducted at the University of Cyprus and the CYENS Centre of Excellence on AI-driven character animation and motion synthesis. We begin by introducing our data acquisition capabilities and an effective motion representation for deep character animation, followed by the introduction of a new motion capture technology for in-the-wild data acquisition that eliminates the need for external cameras and enables robust operation in uncontrolled and outdoor environments. Building on large-scale motion data, this presentation shows how we investigate contextual motion analysis methods that integrate temporal structure, action semantics, and affective cues. A central theme of the talk is the need for an intermediate representation, the so-called movement language model, that bridges natural language and learned motion manifolds, enabling controllable motion generation via generative models with explicit control over sentiment, intent, and task execution. We also present text-prompted motion synthesis pipelines that support compositional and multi-objective generation. Finally, we present our recent work on multimodal generative frameworks for music- and MIDI-driven motion synthesis, aligning rhythmic, harmonic, and temporal features with motion dynamics using deep generative architectures.