论文：DiffTalk: Crafting Diffusion Models for Generalized Audio-Driven Portraits Animation 代码：https://sstzal.github.io/DiffTalk/ 出处：CVPR2023 特点：需要音频+对应人物的视频来合成新的说话头视频，嘴部抖动严重 一、背景 talking head 合成任务

code：GitHub - sstzal/DiffTalk: [CVPR2023] The implementation for "DiffTalk: Crafting Diffusion Models for Generalized Audio-Driven Portraits Animation" 问题1. ERROR: Failed building wheel for pysptk