MarioPD/THVD

这是一个全面的 Talking Head 视频数据集，包含超过50,000个视频，总时长超过500小时，涵盖了来自世界各地的23,841个独特身份。数据集的总大小为2.5TB，包含48,200个视频，涵盖了24,100个身份。视频分辨率为60% 4k (1980p)和33%全高清(1080p)，格式为MP4。每个视频帧都有可见的嘴部运动，最小人脸大小为400像素，视频时长从20秒到5分钟不等。该数据集适用于人脸识别与验证、行为识别、重识别、Deepfake检测、生成式AI、唇同步应用以及背景AI应用等多种场景。数据集覆盖全球范围，包含多种语言的视频，如英语、葡萄牙语、西班牙语等。数据集尚未完成，正在每日扩展。

This is a comprehensive Talking Head Video Dataset containing over 50,000 videos, totaling more than 500 hours of footage and featuring 23,841 unique identities from around the world. The dataset is 2.5TB in size, with 48,200 videos covering 24,100 identities. The resolution is 60% 4k (1980p) and 33% full HD (1080p), with MP4 format. Each video frame has visible mouth movements, a minimum face size of 400 pixels, and video durations ranging from 20 seconds to 5 minutes. The dataset is suitable for a variety of applications including face recognition and verification, action recognition, re-identification, deepfake detection, generative AI, lip syncing applications, and background AI applications. It covers a global scope and includes videos in multiple languages such as English, Portuguese, Spanish, etc. The dataset is not yet complete and is expanding daily.