This paper introduces a novel approach to modelling of volumetric cardiac magnetic resonance imaging (MRI) with simultaneous resolution of cardiac and respiratory motion. The major challenge is that the inherent slow nature of MRI prevents obtaining real-time volumetric images of the heart with sufficient spatial and temporal resolution. To overcome this problem our method predicts pixel intensities in multiple 2D slices, acquired with high spatial and temporal resolution, and subsequently assembles these into volumetric data sets. The prediction is based on external motion sensors, in our case a respiratory bellow and a vectorcardiogram, and utilizes a combination of deformation modelling and pixel intensity modelling. We demonstrate that this approach reliably models volumetric cardiac MRI for any combination of cardiac and respiratory phase.