[Help]: Reconstruct Emilia dataset using Raw data

[3manifold]

Problem Overview

I plan to improve several Emilia pipeline components and reconstruct Emilia dataset using the Emilia raw audio files. Documentation suggests to "download the raw audio files from the provided URL list":

(...)To reconstruct the Emilia dataset, you can download the raw audio files from the provided URL list and use our open-source Emilia-Pipe preprocessing pipeline to process the raw data and rebuild the dataset. Additionally, users can employ Emilia-Pipe to preprocess their own raw speech data to meet specific needs. (...)

Nevertheless, downloading/scraping 101,000 hours of audio from sources similar to YouTube is a daunting challenge. Could you please clarify the procedure for obtaining direct access to the raw audio data?