The DreamVoice dataset is designed to support research in text-guided voice conversion and synthetic voice generation. The dataset is structured into three main files:
dreamvoice_s1_raw_label.csv: Contains raw keyword votes from multiple experts. Each row represents the collective opinions of experts regarding various aspects of the speaker's voice characteristics.
dreamvoice_s2_keyword.csv: Provides verified keywords for each speaker… See the full description on the dataset page: https://huggingface.co/datasets/Higobeatz/DreamVoiceDB.
