toloka/VOX-DUB

VOX-DUB数据集是一个用于评估AI配音系统的基准数据集。它包括来自真实视频的原始语音片段及其相应的翻译文本、由多个配音/TTS系统生成的音频记录、以及人类对五个方面的成对A/B（+ SAME）评估结果：发音、自然度、音质、情感相似度和声音相似度。数据集分为三个主要部分：原始数据、合成数据和注释，每个部分都有特定的功能和用法指南。README还包括对注释者的指导，说明如何根据给定参数评估音频样本。

VOX-DUB is a human-based benchmark for evaluating AI dubbing systems. It includes audio fragments with original speech from real videos and their corresponding translated texts, generated audio recordings produced by multiple dubbing/TTS systems, and human annotation results with pairwise A/B (+ SAME) evaluations across five aspects (pronunciation, naturalness, sound quality, emotion similarity, and voice similarity). The dataset is structured into three main parts: source_data, synthesized_data, and annotations, each with specific features and usage guidelines.