Latest / Kana & Mari's SoundRepos (English)

keonlee9420/STYLER | expressive controllable neural TTS | GitHub
STYLER is a non-autoregressive neural text-to-speech framework for expressive, controllable, rapid, and noise-robust speech synthesis. It models disentangled style factors such as duration, pitch, energy, speaker identity, and noise, supporting supervised training, domain-adversarial noise separation, reference-audio style transfer, and HiFi-GAN vocoding. Related: keonlee9420/STYLER / expressive controllable neural TTS / GitHub
The skinny
The skinny isn't ready yet — notes appear once the transcript is processed.