Latest / Kana & Mari's SoundRepos (English)

zhenye234/FlashSpeech | efficient zero shot speech synthesis | GitHub
FlashSpeech is an implementation of the FlashSpeech research model for efficient zero-shot speech synthesis, adapted from Amphion's NaturalSpeech2. It includes LibriTTS-oriented data preparation, acoustic and phoneme feature extraction, staged model training with PyTorch Lightning, inference, and comprehensive speech-quality evaluation metrics. Related: zhenye234/FlashSpeech / efficient zero shot speech synthesis / GitHub
The skinny
The skinny isn't ready yet — notes appear once the transcript is processed.