arxivcs.CV2026-07-23
GroupVideo: Multi-Identity Customized Text-to-Video Generation
Xinyang Song, Libin Wang, Jianxin Sun, Qi Li, Dandan Zheng, JingDong Chen, et al.
Current identity customized video generation methodologies are predominantly limited to single-identity scenarios, as the lack of explicit identity separation mechanisms often leads to identity confusion in multi-identity settings. Existing multi-identity approaches, which direct…