FENRlR/MB-iSTFT-VITS2

Application of MB-iSTFT-VITS components to vits2_pytorch

52
/ 100
Established

Combines multi-band inverse Short-Time Fourier Transform (MB-iSTFT) vocoding with VITS2's end-to-end text-to-speech architecture, enabling subband-wise synthesis for improved audio quality. Supports multiple alignment backends including Triton-accelerated Super Monotonic Align, eliminating Cython compilation requirements. Offers variants ranging from full MB-iSTFT-VITS2 to lightweight Mini configurations, with single and multi-speaker training pipelines.

134 stars.

No Package No Dependents
Maintenance 6 / 25
Adoption 10 / 25
Maturity 16 / 25
Community 20 / 25

How are scores calculated?

Stars

134

Forks

31

Language

Python

License

MIT

Last pushed

Dec 29, 2025

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/voice-ai/FENRlR/MB-iSTFT-VITS2"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.