Soul-Bench is an evaluation benchmark for audio-driven human animation tasks, released by Tencent YouTu Lab in 2025. Related research papers include... Soul: Breathe Life into Digital Human for High-fidelity Long-term Multimodal AnimationThe aim is to systematically evaluate the generation quality, consistency, and generalization ability of relevant methods in real-world application scenarios. This dataset contains 226 video test samples, exhibiting a relatively rich distribution across multiple dimensions, as detailed below:

• Main body type distribution

• Upper body scenes: 107
• Full-body scenes: 72
• Portraits, animated characters, and animals: 47 items

• Audio type distribution

• Dialogue-based audio: 177 pieces
• Vocal performances: 49 items

• Video resolution distribution

• 1080P: 118 items
• 720P: 55 items
• 4K: 51 items
• 480P: 2 items

• Screen proportions

• 1 < r ≤ 2: 170 entries
• r = 1 (square): 44 lines
• 0.5 ≤ r < 1 (vertical): 12 lines

• Video duration distribution

• 27–30 second interval: 70 lines

  

Citation

@misc{soul,
title={Soul: Breathe Life into Digital Human for High-fidelity Long-term Multimodal Animation},
author={Jiangning Zhang and Junwei Zhu and Zhenye Gan and Donghao Luo and Chuming Lin and Feifan Xu and Xu Peng and Jianlong Hu and Yuansen Liu and Yijia Hong and Weijian Cao and Han Feng and Xu Chen and Chencan Fu and Keke He and Xiaobin Hu and Chengjie Wang},
year={2025},
eprint={2512.13495},
archivePrefix={arXiv},
primaryClass={cs.CV},
url={https://arxiv.org/abs/2512.13495},
}