|
import os |
|
from extra import TTSTokenizer, VitsConfig, CharactersConfig, VitsCharacters |
|
import torch |
|
import numpy as np |
|
|
|
|
|
letters="खछगचऊुलणशढ़इौज़झठैढजफ़औ्ड़फूेानटॅयव़ऋदप.थअँऑआघहतषरसभउञडएईऐक़ िओ?धी,ॉंख़कोबमृ" |
|
model="ch_male_vits_30hrs.pt" |
|
text = "पेरिविंकल के जड़, उपजी अउ पत्त्ता मन ह बिकट उपयोगी हे" |
|
|
|
config = VitsConfig( |
|
text_cleaner="multilingual_cleaners", |
|
characters=CharactersConfig( |
|
characters_class=VitsCharacters, |
|
pad="<PAD>", |
|
eos="<EOS>", |
|
bos="<BOS>", |
|
blank="<BLNK>", |
|
characters=letters, |
|
punctuations="!¡'(),-.:;¿? ", |
|
phonemes=None) |
|
) |
|
tokenizer, config = TTSTokenizer.init_from_config(config) |
|
|
|
x = tokenizer.text_to_ids(text) |
|
x = torch.from_numpy(np.array(x)).unsqueeze(0) |
|
net = torch.jit.load(model) |
|
with torch.no_grad(): |
|
out2 = net(x) |
|
import soundfile as sf |
|
sf.write("jit.wav", out2.squeeze().cpu().numpy(), 22050) |
