Who is a Better Talker (This repo is now not available, still the same as [Imitator](https://github.com/zyj-2000/Imitator)

Yingjie Zhou Zicheng Zhang, Jun Jia, Yanwei Jiang, Xiaohong Liu, Xiongkuo Min and Guangtao Zhai

Email: zyj2000@sjtu.edu.cn

---

Introduction

This repository explores the fascinating question of imitation capabilities across different models. We investigate various approaches to talking head generation and evaluate their effectiveness in replicating complex talking heads. Our work provides insights into the fundamental differences between imitation learning paradigms.

---

Visualization

Figure 1: Comparative visualization of different imitation approaches.

---

P2P-Motion

Our novel Peer-to-Peer Motion (P2P-Motion) framework establishes a new benchmark in imitation learning. Key features include:

• Real-time motion transfer between heterogeneous agents
• Adaptive spatial-temporal alignment
• Bidirectional imitation capability
• Robustness to morphological differences

---

Getting Started (Attention: Not THQA-10K now)

To use this Dataset (Not THQA-10K now):

百度云盘：AHQA_Dataset

To use sources in Dataset (including Designed Animation and Generated Human Images):

百度云盘：AHQA_source

git clone https://github.com/yourusername/who-is-a-better-imitator.git
cd who-is-a-better-imitator
pip install -r requirements.txt