candlewill/Speech-Corpus-Collection

Curates publicly available speech datasets across multiple scales and languages, including large-scale corpora (LibriSpeech's 1000 hours), specialized domains (TED talks), and single-speaker databases optimized for voice synthesis. Provides centralized access to diverse ASR training data like VCTK and TEDLIUM alongside TTS corpora from CMU ARCTIC and Blizzard Challenge resources, with preprocessing notes for datasets requiring manual alignment like the World English Bible.

113 stars. No commits in the last 6 months.