HyperAI

FakeAVCeleb Audio and Video Deepfake Dataset

Date

2 years ago

Organization

Publish URL

sites.google.com

License

CC BY 4.0

Categories

Download Help
特色图像

FakeAVCeleb is an audio-video Deepfake dataset that contains both fake videos and the corresponding synthetic cloned audio.

This dataset contains 500 videos with an average duration of 7.8 seconds selected from the VoxCeleb2 dataset, one video for each celebrity. This dataset aims to select videos equally based on gender, race, and age. Each race contains 100 videos of 100 celebrities. The ratio of males to females for each race is 50%, that is, 50 male videos and 50 female videos out of 100 videos.