Metadata-Version: 2.5
Name: sanotts
Version: 0.3.0
Summary: Self-contained, numpy-only inference runtime for sanoTTS voices, including the 294k and 2.27M heart voices.
License: GPL-3.0-or-later
Requires-Python: >=3.10
Requires-Dist: espeakng-loader<1,>=0.2
Requires-Dist: numpy>=1.24
Requires-Dist: phonemizer-fork<4,>=3.3
Description-Content-Type: text/plain

sanotts: pip-installable, numpy-only inference for sanoTTS voices. Two graphs are supported: the piperlite voices (amy, kristin, hfc, vi, id) and the nano voices (heart at 2.27M parameters, heart-nano at 294k), the latter running mel-100 -> ConvNeXt1D -> iSTFT with a noise-fed decoder. Both are gated against the same PyTorch references the C runtime is held to. Voices download on demand or load from a local directory.