As a new trend of AIGC, zero-shot speech synthesis has wide-ranging applications, including voice assistants, audiobooks, video game character voices, creating podcast, and real-time voice changer, among others. Magic Data has proactively designed and developed the "High-Quality Dataset for Voice Replication," which boasts a 48kHz high sampling rate, tens of thousands of contributors with diverse content, making it an excellent resource for achieving zero-shot speech synthesis. In particular, the diverse content of this data can be beneficial for the model's coarticulation learning.