请问效果咋样呢？ about adaspeech HOT 9 CLOSED

Sorry, but, i don't understand chinese.
If you don't mind, can you use english?

from adaspeech.

sorry for the chinese
May I ask how you reproduced this project? Can the voice be customized? It can be reproduced by inputting one or two hundred voices by one person.
and can you release your pretrained model? Thx

from adaspeech.

Sorry, currently, i'm not planning to release adaptation code.
and, i upload the pretrained model using LJspeech and Kokoro dataset.

from adaspeech.

You mean that this code cannot custom voice?
so what the difference be this code and FastSpeech2?
and where is the pretrained model using LJspeech and Kokoro dataset?

from adaspeech.

My focus was to improve FastSpeech 2 acoustic generalization using Utterance level encoder and Phoneme level encoder rather than adaptation. So, adaptation was not my focus.

And, you can find pretrained network here, this is same link in readme.md

from adaspeech.

could you please share your log file? i want to know after your training, the value of your loss

from adaspeech.

and i want train a mandarin model , what should i change except symbols?

from adaspeech.

1. Give me your email address. Then, i will send you LJspeech training log file.
2. You need to change init.py, symbols.py and cleaners.py in text folder. My code is based on ming's code that support mandarin. Checking this codes might help.
3. You need to make preprocessing step yourself, check preprocessor folder's ljspeech.py or other .py files

from adaspeech.

[email protected]
thank you very much

from adaspeech.