problems about hifigan model

[asr-pub]

Hello, I am sorry to disturb you.
I want to train a hifigan model, so I use the default train script "TTS/recipes/ljspeech/hifigan/train_hifigan.py", and didn't modify anything.
When I train about 200K steps, I open tensorboard and listen the synthesis audio, The audio doesn't sound right, here is the audio link. audio
Have I missed something?

[Sadam1195]

Maybe your audio parameters of source wav files and config file don't compliment each other very well? @asr-pub

[Sadam1195]

try feat_match_loss_weight = 200 as recommended in this thread #677 (reply in thread)

[asr-pub]

OK, thank you, I try it now.

[cdlIris]

hi, I want to train my own hifigan, and encountered the same problem as you.Do you have any updates after the modification? thx

[Sadam1195]

hi, I want to train my own hifigan, and encountered the same problem as you.Do you have any updates after the modification? thx

please define "same" .
@cdlIris

[asr-pub]

hi, I want to train my own hifigan, and encountered the same problem as you.Do you have any updates after the modification? thx

Recently, I use nvidia-tacotron2 and official hifigan , then I get a nice result.