tuanh123789/Train_Hifigan_XTTS

Training HiFi-GAN to Upsample from 22,050 Hz to 48,000 Hz

Opened this issue · 1 comments

Hello,

I would like to ask if it's possible to train the HiFi-GAN model to upsample from 22,050 Hz directly to 48,000 Hz instead of the default 24,000 Hz.

I'm planning to make the following changes:

Use a dataset with 48 kHz high-quality audio samples for training.
Adjust the model configuration (e.g., set target_sample_rate to 48,000 Hz).
Are there any other important considerations or adjustments I should be aware of when training the model for this higher upsampling rate? Any advice or suggestions would be greatly appreciated!

Thank you!

To train 48k you need to change the dvae encoding. Almost had to retrain the entire model