fishaudio / Bert-VITS2

vits2 backbone with multilingual-bert

wavtechyukky / LeapSinger

Fast singing-voice acoustic model: harmonic + noise excitation refined into a mel by a 1-step rectified flow (bundled NHVSing vocoder). Multi-speaker + style control; ONNX/OpenUTAU export.