Spaces:
Running
Running
| from onnx_modules.V230_OnnxInference import OnnxInferenceSession | |
| import numpy as np | |
| import torch | |
| from scipy.io.wavfile import write | |
| from text import cleaned_text_to_sequence, get_bert | |
| from text.cleaner import clean_text | |
| import utils | |
| import commons | |
| hps = utils.get_hparams_from_file('onnx/BangDreamApi.json') | |
| device = 'cpu' | |
| Session = OnnxInferenceSession( | |
| { | |
| "enc" : "onnx/BangDreamApi/BangDreamApi_enc_p.onnx", | |
| "emb_g" : "onnx/BangDreamApi/BangDreamApi_emb.onnx", | |
| "dp" : "onnx/BangDreamApi/BangDreamApi_dp.onnx", | |
| "sdp" : "onnx/BangDreamApi/BangDreamApi_sdp.onnx", | |
| "flow" : "onnx/BangDreamApi/BangDreamApi_flow.onnx", | |
| "dec" : "onnx/BangDreamApi/BangDreamApi_dec.onnx" | |
| }, | |
| Providers = ["CPUExecutionProvider"] | |
| ) | |
| def get_text(text, language_str, hps, device, style_text=None, style_weight=0.7): | |
| style_text = None if style_text == "" else style_text | |
| norm_text, phone, tone, word2ph = clean_text(text, language_str) | |
| phone, tone, language = cleaned_text_to_sequence(phone, tone, language_str) | |
| if True: | |
| phone = commons.intersperse(phone, 0) | |
| tone = commons.intersperse(tone, 0) | |
| language = commons.intersperse(language, 0) | |
| for i in range(len(word2ph)): | |
| word2ph[i] = word2ph[i] * 2 | |
| word2ph[0] += 1 | |
| bert_ori = get_bert( | |
| norm_text, word2ph, language_str, device, style_text, style_weight | |
| ) | |
| del word2ph | |
| assert bert_ori.shape[-1] == len(phone), phone | |
| if language_str == "ZH": | |
| bert = bert_ori | |
| ja_bert = torch.randn(1024, len(phone)) | |
| en_bert = torch.randn(1024, len(phone)) | |
| elif language_str == "JP": | |
| bert = torch.randn(1024, len(phone)) | |
| ja_bert = bert_ori | |
| en_bert = torch.randn(1024, len(phone)) | |
| elif language_str == "EN": | |
| bert = torch.randn(1024, len(phone)) | |
| ja_bert = torch.randn(1024, len(phone)) | |
| en_bert = bert_ori | |
| else: | |
| raise ValueError("language_str should be ZH, JP or EN") | |
| assert bert.shape[-1] == len( | |
| phone | |
| ), f"Bert seq len {bert.shape[-1]} != {len(phone)}" | |
| phone = torch.LongTensor(phone) | |
| tone = torch.LongTensor(tone) | |
| language = torch.LongTensor(language) | |
| return bert, ja_bert, en_bert, phone, tone, language | |
| def infer( | |
| text, | |
| sid, | |
| style_text=None, | |
| style_weight=0.7, | |
| sdp_ratio=0.5, | |
| noise_scale=0.6, | |
| noise_scale_w=0.667, | |
| length_scale=1, | |
| ): | |
| language= 'JP' if is_japanese(text) else 'ZH' | |
| bert, ja_bert, en_bert, phones, tone, language = get_text( | |
| text, | |
| language, | |
| hps, | |
| device, | |
| style_text=style_text, | |
| style_weight=style_weight, | |
| ) | |
| with torch.no_grad(): | |
| x_tst = phones.unsqueeze(0).to(device).numpy() | |
| language = np.zeros_like(x_tst) | |
| tone = np.zeros_like(x_tst) | |
| bert = bert.to(device).transpose(0, 1).numpy() | |
| ja_bert = ja_bert.to(device).transpose(0, 1).numpy() | |
| en_bert = en_bert.to(device).transpose(0, 1).numpy() | |
| del phones | |
| sid = np.array([hps.spk2id[sid]]) | |
| audio = Session( | |
| x_tst, | |
| tone, | |
| language, | |
| bert, | |
| ja_bert, | |
| en_bert, | |
| sid, | |
| seed=114514, | |
| seq_noise_scale=noise_scale_w, | |
| sdp_noise_scale=noise_scale, | |
| length_scale=length_scale, | |
| sdp_ratio=sdp_ratio, | |
| ) | |
| del x_tst, tone, language, bert, ja_bert, en_bert, sid | |
| write('temp.wav', 44100, audio) | |
| def is_japanese(string): | |
| for ch in string: | |
| if ord(ch) > 0x3040 and ord(ch) < 0x30FF: | |
| return True | |
| return False | |
| if __name__ == "__main__": | |
| infer("你好，我是说的道理", "パレオ") | |
| ''' | |
| from onnx_modules.V230_OnnxInference import OnnxInferenceSession | |
| import numpy as np | |
| import torch | |
| from scipy.io.wavfile import write | |
| from text import cleaned_text_to_sequence, get_bert | |
| from text.cleaner import clean_text | |
| import utils | |
| import commons | |
| hps = utils.get_hparams_from_file('onnx/BangDreamApi.json') | |
| device = 'cpu' | |
| Session = OnnxInferenceSession( | |
| { | |
| "enc" : "onnx/BangDreamApi/BangDreamApi_enc_p.onnx", | |
| "emb_g" : "onnx/BangDreamApi/BangDreamApi_emb.onnx", | |
| "dp" : "onnx/BangDreamApi/BangDreamApi_dp.onnx", | |
| "sdp" : "onnx/BangDreamApi/BangDreamApi_sdp.onnx", | |
| "flow" : "onnx/BangDreamApi/BangDreamApi_flow.onnx", | |
| "dec" : "onnx/BangDreamApi/BangDreamApi_dec.onnx" | |
| }, | |
| Providers = ["CPUExecutionProvider"] | |
| ) | |
| def get_text(text, language_str, hps, device, style_text=None, style_weight=0.7): | |
| style_text = None if style_text == "" else style_text | |
| norm_text, phone, tone, word2ph = clean_text(text, language_str) | |
| phone, tone, language = cleaned_text_to_sequence(phone, tone, language_str) | |
| if True: | |
| phone = commons.intersperse(phone, 0) | |
| tone = commons.intersperse(tone, 0) | |
| language = commons.intersperse(language, 0) | |
| for i in range(len(word2ph)): | |
| word2ph[i] = word2ph[i] * 2 | |
| word2ph[0] += 1 | |
| bert_ori = get_bert( | |
| norm_text, word2ph, language_str, device, style_text, style_weight | |
| ) | |
| del word2ph | |
| assert bert_ori.shape[-1] == len(phone), phone | |
| if language_str == "ZH": | |
| bert = bert_ori | |
| ja_bert = torch.randn(1024, len(phone)) | |
| en_bert = torch.randn(1024, len(phone)) | |
| elif language_str == "JP": | |
| bert = torch.randn(1024, len(phone)) | |
| ja_bert = bert_ori | |
| en_bert = torch.randn(1024, len(phone)) | |
| elif language_str == "EN": | |
| bert = torch.randn(1024, len(phone)) | |
| ja_bert = torch.randn(1024, len(phone)) | |
| en_bert = bert_ori | |
| else: | |
| raise ValueError("language_str should be ZH, JP or EN") | |
| assert bert.shape[-1] == len( | |
| phone | |
| ), f"Bert seq len {bert.shape[-1]} != {len(phone)}" | |
| phone = torch.LongTensor(phone) | |
| tone = torch.LongTensor(tone) | |
| language = torch.LongTensor(language) | |
| return bert, ja_bert, en_bert, phone, tone, language | |
| def infer( | |
| text, | |
| sid, | |
| style_text=None, | |
| style_weight=0.7, | |
| ): | |
| language= 'JP' if is_japanese(text) else 'ZH' | |
| bert, ja_bert, en_bert, phones, tone, language = get_text( | |
| text, | |
| language, | |
| hps, | |
| "cpu", | |
| style_text=style_text, | |
| style_weight=style_weight, | |
| ) | |
| with torch.no_grad(): | |
| x_tst = phones.unsqueeze(0).to(device).numpy() | |
| tone = tone.to(device).unsqueeze(0).numpy() | |
| bert = bert.to(device).transpose(0, 1).numpy() | |
| ja_bert = ja_bert.to(device).transpose(0, 1).numpy() | |
| en_bert = en_bert.to(device).transpose(0, 1).numpy() | |
| del phones | |
| language = np.zeros_like(x_tst) | |
| tone = np.zeros_like(x_tst) | |
| print(bert) | |
| print(tone) | |
| print(ja_bert) | |
| print(language) | |
| sid = np.array([0]) | |
| audio = Session( | |
| x_tst, | |
| tone, | |
| language, | |
| bert, | |
| ja_bert, | |
| en_bert, | |
| sid | |
| ) | |
| write('temp.wav', 44100, audio) | |
| def is_japanese(string): | |
| for ch in string: | |
| if ord(ch) > 0x3040 and ord(ch) < 0x30FF: | |
| return True | |
| return False | |
| if __name__ == "__main__": | |
| infer("你好，我是说的道理", "香澄") | |
| ''' |