The paper introduces the Fréchet Video Motion Distance (FVMD), a novel metric designed to evaluate the temporal consistency and motion quality of generated videos. Unlike existing metrics like FID-VID and FVD, which focus on visual quality and temporal coherence, respectively, FVMD specifically addresses the challenge of ensuring that generated videos maintain consistent and realistic motion patterns. The authors propose extracting motion features from key point trajectories using a pre-trained model (PIPs++) and then measuring the similarity between these features using the Fréchet distance. Extensive experiments, including sensitivity analysis and human studies, demonstrate that FVMD effectively captures temporal noise and aligns better with human perceptions of video quality compared to existing metrics. Additionally, the motion features derived from FVMD consistently improve the performance of Video Quality Assessment (VQA) models, highlighting its potential for unary video quality evaluation tasks. The code for FVMD is available at <https://github.com/ljh0v0/FMD-frechet-motion-distance>.The paper introduces the Fréchet Video Motion Distance (FVMD), a novel metric designed to evaluate the temporal consistency and motion quality of generated videos. Unlike existing metrics like FID-VID and FVD, which focus on visual quality and temporal coherence, respectively, FVMD specifically addresses the challenge of ensuring that generated videos maintain consistent and realistic motion patterns. The authors propose extracting motion features from key point trajectories using a pre-trained model (PIPs++) and then measuring the similarity between these features using the Fréchet distance. Extensive experiments, including sensitivity analysis and human studies, demonstrate that FVMD effectively captures temporal noise and aligns better with human perceptions of video quality compared to existing metrics. Additionally, the motion features derived from FVMD consistently improve the performance of Video Quality Assessment (VQA) models, highlighting its potential for unary video quality evaluation tasks. The code for FVMD is available at <https://github.com/ljh0v0/FMD-frechet-motion-distance>.