VoxelMorph is a fast, learning-based framework for deformable medical image registration. Unlike traditional methods that optimize an objective function for each pair of images, VoxelMorph formulates registration as a function that maps an input image pair to a deformation field. This function is parameterized by a convolutional neural network (CNN) and optimized using a training set of images. For new pairs of scans, VoxelMorph rapidly computes a deformation field by evaluating the learned function. The paper explores two training strategies: an unsupervised approach that maximizes standard image matching objectives and a supervised approach that leverages available segmentations. The unsupervised model achieves comparable accuracy to state-of-the-art methods but operates orders of magnitude faster. VoxelMorph trained with auxiliary data improves registration accuracy at test time, and the effect of training set size on registration is evaluated. The method promises to speed up medical image analysis and processing pipelines while facilitating novel directions in learning-based registration.VoxelMorph is a fast, learning-based framework for deformable medical image registration. Unlike traditional methods that optimize an objective function for each pair of images, VoxelMorph formulates registration as a function that maps an input image pair to a deformation field. This function is parameterized by a convolutional neural network (CNN) and optimized using a training set of images. For new pairs of scans, VoxelMorph rapidly computes a deformation field by evaluating the learned function. The paper explores two training strategies: an unsupervised approach that maximizes standard image matching objectives and a supervised approach that leverages available segmentations. The unsupervised model achieves comparable accuracy to state-of-the-art methods but operates orders of magnitude faster. VoxelMorph trained with auxiliary data improves registration accuracy at test time, and the effect of training set size on registration is evaluated. The method promises to speed up medical image analysis and processing pipelines while facilitating novel directions in learning-based registration.