MegActor generates vivid synthetic videos from raw video content

Tutorial Introduction

This tutorial is a demo of MegActor. You can experience it by cloning and opening the API address. MegActor is a portrait animation generator developed by MEGVII Technology that uses raw videos as a driving factor to generate realistic and vivid talking head videos. Specifically, the research team used two unets: one to extract identity and background features from the source image, and the other to accurately generate and integrate motion features extracted directly from the raw video. MegActor can be trained on low-quality, publicly available datasets and excels in facial expressions, pose diversity, subtle controllability, and visual quality.

This technology belongs to the field of intermediate-representation-free portrait animation, which can directly generate realistic and vivid head-view videos using raw videos as driving factors. The core advantage of MegActor is that it abandons the representation form used as an intermediary in traditional animation production and directly uses raw video data to create animation, which greatly improves the efficiency and quality of animation production.

Run steps

---

Step 1: Resource Configuration

在该项目右上角点击「克隆」，随后依次点击 “下一步” 即可完成：基本信息> 选择算力> 审核等步骤。最后点击「继续执行」即可在个人容器内开启本项目。

Step 2: Project Startup

待资源分配完成后，参照下图中的步骤启动项目。

Step 3: Video Generation

上传图片文件和视频文件进行合成

(The synthesis time is about 3 minutes)

Right click the resulting video to save it locally

Discussion and Exchange

🖌️ If you see a high-quality project, please leave a message in the background to recommend it! In addition, we have also established a tutorial exchange group. Welcome friends to scan the QR code and remark [SD Tutorial] to join the group to discuss various technical issues and share application effects↓