The paper introduces a new optical flow dataset, MPI-Sintel, derived from the open-source 3D animated short film *Sintel*. This dataset addresses the limitations of existing datasets like Middlebury, which lack realistic motion, complexity, and diversity. MPI-Sintel includes long sequences, large motions, specular reflections, motion blur, defocus blur, and atmospheric effects. The dataset is publicly available, and the authors evaluate several recent optical flow algorithms, finding that top methods on Middlebury perform poorly on MPI-Sintel, highlighting the need for further research. The paper also compares the image and flow statistics of MPI-Sintel to those of real films and videos, showing that they are similar, validating the use of synthetic data. The dataset, metrics, and evaluation website are provided to facilitate future research and comparisons.The paper introduces a new optical flow dataset, MPI-Sintel, derived from the open-source 3D animated short film *Sintel*. This dataset addresses the limitations of existing datasets like Middlebury, which lack realistic motion, complexity, and diversity. MPI-Sintel includes long sequences, large motions, specular reflections, motion blur, defocus blur, and atmospheric effects. The dataset is publicly available, and the authors evaluate several recent optical flow algorithms, finding that top methods on Middlebury perform poorly on MPI-Sintel, highlighting the need for further research. The paper also compares the image and flow statistics of MPI-Sintel to those of real films and videos, showing that they are similar, validating the use of synthetic data. The dataset, metrics, and evaluation website are provided to facilitate future research and comparisons.