PCA on all 54 official Kokoro voicepacks Projected hf_alpha into PCA space Shifted individual PCs and combinations Hindi phonetics preserved (all start from hf_alpha)

Files

*.npy — Voicepack tensors [510, 1, 256] float32 *.ogg — Sample audio (Hindi text) catalog.json — Metadata

	Single PC shifts (60): pc{1-10}_{±5,±15,±25}… See the full description on the dataset page: https://huggingface.co/datasets/hf4uwho/kokoro-hindi-voice-variations.