Multimedia is ubiquitous in 21st-century education. Cognitive Load Theory and the Cognitive Theory of Multimedia Learning both postulate that the quality of multimedia design heavily influences learning. We sought to identify how to best design multimedia, and review how well those learning theories held up to meta-analyses. We conducted an overview of systematic reviews that tested the effects of multimedia design on learning or cognitive load. We found 29 reviews including 1,189 studies and 78,177 participants. We found 11 design principles that demonstrated significant, positive, meta-analytic effects on learning, and five that significantly improved management of cognitive load. The largest benefits were for captioning second-language videos, temporal/spatial contiguity, and signaling. We also found robust evidence for modality, animation, coherence/removing seductive details, anthropomorphics, segmentation, personalisation, pedagogical agents, and verbal redundancy effects. Good design was more important for more complex materials, and in system-paced environments (e.g., lectures) than self-paced ones (e.g., websites). Results supported many tenets of both theories. We highlight a range of evidence-based strategies that could be implemented by educators.