Sellopale/Soul-Bench

Soul数据集是一个专注于数字人类高保真长期多模态动画的数据集，支持图像到视频转换、表格问答和翻译等任务。它包含动画、音频、图像和视频等多种数据类型，适用于非商业用途的研究和开发。

The Soul dataset is focused on high-fidelity long-term multimodal animation for digital humans, supporting tasks such as image-to-video conversion, table question answering, and translation. It includes various data types like animation, audio, image, and video, and is intended for non-commercial research and development purposes.