Intro

This dataset will help in training the Speaker encoder of a speaker style task. Read more about it here

Read ME

This notebook will teach you to download audio data from the Voxceleb1 database.

Couple of things to note The original audio data is divided into multiple files on the Voxceleb website, you can find more details here We will use DevA as a sample, but you can scale it as per your convinent

Steps to get your customer or all of the VoxCeleb data (just follow the steps here)

• Download the data using wget (depending upon the disk space, you can download all or just one)
• One part file is of 10GB size
• Merge part files into one '.zip' file
• Extract the zip file into normal 'wav' files
• Use a function to move desired celebrity's wav files to another folder
• zip the final folder and download and use

Acknowledgements

A huge thank you to Nagrani to prepare the huge dataset

A. Nagrani*, J. S. Chung*, A. Zisserman VoxCeleb: a large-scale speaker identification dataset INTERSPEECH, 2017. Bibtex | Abstract | PDF | Presentation

Inspiration

Questions is, how and can you use this data to start style transfer in audio, for more details click here