search

Rayhane-mamah / Tacotron-2

DeepMind's Tacotron-2 Tensorflow implementation
MIT License
2.25k stars 911 forks source link

Preprocess problem #510

Open chalathecfc opened 3 years ago

chalathecfc commented 3 years ago

after preprocessing a dataset that has similar structure with that of LJspeech only 7731 of 8076 audios files are processed i don't know where the problem is from. how can i fix this problem? any idea?

pandaGst commented 2 years ago

Maybe they have been droped because they are too long, the number of frames of them is larger than max_mel_frames in hparams.py#70. You can also see the using of this param in datasets/preprocessor.py