The recent developments in human motion generation and editing have significantly advanced the field, focusing on enhancing the quality, precision, and applicability of generated motions. A notable trend is the integration of large language models and vision models to create more sophisticated and context-rich motion sequences. This approach allows for detailed text-guided motion editing and self-correction, leading to improved performance in complex and unseen motions. Additionally, the use of hierarchical models and residual vector quantization has enabled the generation of longer, more lifelike motion sequences, addressing the limitations of existing datasets. The decoupling of upper-body control from locomotion in humanoid robots, combined with predictive motion priors, has also shown promise in achieving robust whole-body control for precise manipulation tasks. Notably, the introduction of novel architectures like Mogo, which leverages a hierarchical causal transformer, has set new benchmarks in high-quality 3D human motion generation, particularly in out-of-distribution scenarios. These advancements collectively push the boundaries of what is possible in human motion modeling and its applications across various domains.