Bangla Audio Dataset: Original and DeepFake Voices for AI-Based Voice Analysis and Detection
Published: 18 January 2024| Version 1 | DOI: 10.17632/4ftmwt86vr.1
Contributors:
Md Akteruzzaman Dipto, , Description
Total number of audios : 4500 (Original: 2250; Fake: 2250) Total number of individuals: 75 Total number of distinct sets: 15 Total number of distinct sentences per set : 30 Each set spoken by the number of individuals: 5 Each sentence voice time: 2 to 5 seconds Male-Female ratio: Nearly Balanced It has a parent folder named ‘DATASET’ under which there are 75 folders with unique names to identify the characteristics, and under this there are two folders named ‘Real’ and 'Fake', which include 30 voices each named numerically.
Files
Institutions
University of Asia Pacific
Categories
Audio Analysis, Deepfake