v0.4.1

What's Changed

• v0.3.3: train acoustic model on TPU in #18
• v0.4.1: work with the latest MFA format in #22

Full Changelog: v0.3.1...v0.4.1

Add Fine-tune training for HiFi-GAN

Improve audio quality with fine-tuned HiFi-GAN.

v0.3.0

Add word separator token!

Set silence audio segments to zero to reduce noise.

Use the nvidia/tacotron2/prenet with two dropout layers.

v0.2.0

Use HiFiGAN vocoder for faster and better vocoder.

v0.1.2

Update waveRNN pretrained model.

v0.1

We have a working quick start script.

v0.1.1

Model training is now working correctly.