APATE VCD Deepfakes

This dataset was funded by the ANR project APATE: ANR-22-CE39-0016 APATE VCD Deepfakes was constructed from 40 mobile phone selfie videos from https://github.com/microsoft/VCD, i.e. the videos in the "th-m" sub-folder of the VCD dataset. Note that the original videos are not provided with this dataset of deepfake videos. Twenty online deepfake methods were used - 14 face swap and 6 avatar - to create a total of 641 videos; an average of 32 per method. Deepfake methods Face-swap methods Method Num. videos URL AI_Face_Swap_premium 38 https://aifaceswap.io/ AKOOL_FaceSwap 37 https://akool.com/ BestFaceSwap 17 https://www.bestfaceswap.net/ DeepAny 30 https://deepany.ai/ DeepFacer 12 https://deepfacer.com/ DeepSwapper 33 https://www.deepswapper.com/ KreadoAI_FaceSwap 37 https://www.kreadoai.com/ Live3d 30 https://live3d.io/ Magic_Hour_premium 35 https://magichour.ai/ PixNova_premium 37 https://pixnova.ai/ Remaker 30 https://remaker.ai/ RemixAI 36 https://www.remixai.io/ SwapFans 23 https://www.swapfans.ai/ Vidqu 29 https://www.vidqu.ai/ Avatar methods Method Num. videos URL AKOOL_Avatar 40 https://akool.com/ AKOOL_Avatar_audio 40 https://akool.com/ HeyGen 40 https://app.heygen.com/home KreadoAI_Avatar1 40 https://www.kreadoai.com/ KreadoAI_Avatar2 37 https://www.kreadoai.com/ PixNova_Avatar 20 https://pixnova.ai/ LICENCE This dataset was constructed from the Microsoft VCD dataset and is therefore also released under the Community Data License Agreement – Permissive, Version 2.0 - CDLA. See the LICENSE-DATA file for details.