使用自己的声音做语音合成本地部署失败

[ANyyJS]

参考https://zhuanlan.zhihu.com/p/587765776代码就改了下文件路径，运行报错，python和paddle都装好了。 报错：

Traceback (most recent call last): File "E:\Record\anntest\test_1.py", line 26, in <module> frontend = get_frontend( File "D:\Program Files\Python39\lib\site-packages\paddlespeech\t2s\exps\syn_utils.py", line 272, in get_frontend frontend = MixFrontend( File "D:\Program Files\Python39\lib\site-packages\paddlespeech\t2s\frontend\mix_frontend.py", line 29, in __init__ self.zh_frontend = Frontend( File "D:\Program Files\Python39\lib\site-packages\paddlespeech\t2s\frontend\zh_frontend.py", line 147, in __init__ with open(phone_vocab_path, 'rt') as f: OSError: [Errno 22] Invalid argument: 'E:\\Record\x07nntest\\demo\\phone_id_map.txt'

文件夹路径是E:\Record\anntest，模型在E:\Record\anntest\demo，代码如下： `from pathlib import Path import soundfile as sf import os from paddlespeech.t2s.exps.syn_utils import get_am_output from paddlespeech.t2s.exps.syn_utils import get_frontend from paddlespeech.t2s.exps.syn_utils import get_predictor from paddlespeech.t2s.exps.syn_utils import get_voc_output

在其他环境中，记得修改下面这两个变量的路径

am_inference_dir = "E:\Record\anntest\demo" voc_inference_dir = "E:\Record\anntest\pwgan_aishell3_static_1.1.0" # 这里以 pwgan_aishell3 为例子

音频生成的路径，修改成你音频想要保存的路径

wav_output_dir = "E:\Record\anntest\output"

选择设备[gpu / cpu]，这里以GPU为例子，

device = "cpu"

想要生成的文本和对应文件名

text_dict = { "1": "今天天气真不错，欢迎和我一起玩。", "2": "我认为跑步给我的身体带来了健康。", }

frontend

frontend = get_frontend( lang="mix", phones_dict=os.path.join(am_inference_dir, "phone_id_map.txt"), tones_dict=None )

am_predictor

am_predictor = get_predictor( model_dir=am_inference_dir, model_file="fastspeech2_mix" + ".pdmodel", params_file="fastspeech2_mix" + ".pdiparams", device=device)

voc_predictor

voc_predictor = get_predictor( model_dir=voc_inference_dir, model_file="pwgan_aishell3" + ".pdmodel", # 这里以 pwgan_aishell3 为例子，其它模型记得修改此处模型名称 params_file="pwgan_aishell3" + ".pdiparams", device=device)

output_dir = Path(wav_output_dir) output_dir.mkdir(parents=True, exist_ok=True)

sentences = list(text_dict.items())

merge_sentences = True fs = 24000 for utt_id, sentence in sentences: am_output_data = get_am_output( input=sentence, am_predictor=am_predictor, am="fastspeech2_mix", frontend=frontend, lang="mix", merge_sentences=merge_sentences, speaker_dict=os.path.join(am_inference_dir, "phone_id_map.txt"), spk_id=0, ) wav = get_voc_output( voc_predictor=voc_predictor, input=am_output_data)

保存文件

sf.write(output_dir / (utt_id + ".wav"), wav, samplerate=fs)

`

参考用了另一种代码，报了另一种错。 代码： `from pathlib import Path

import soundfile as sf

import os

from paddlespeech.t2s.exps.syn_utils import get_am_output

from paddlespeech.t2s.exps.syn_utils import get_frontend

from paddlespeech.t2s.exps.syn_utils import get_predictor

from paddlespeech.t2s.exps.syn_utils import get_voc_output

def get_text_dict(name:str,txtname:str):

ff = open(txtname,"r",encoding="ISO-8859-1")

msg = ff.read()

ff.close()

text_list = msg.split("\n")

text_dict = {}

num = 0

for i in text_list:

text_dict[name+str(num)] = i

num+=1

print(f"{name}text:{num}")

return text_dict

def the_main(text_dict):

frontend

frontend = get_frontend(

lang="mix",

phones_dict=os.path.join(am_inference_dir, "phone_id_map.txt"),

tones_dict=None

)

am_predictor

am_predictor = get_predictor(

model_dir=am_inference_dir,

model_file="fastspeech2_mix" + ".pdmodel",

params_file="fastspeech2_mix" + ".pdiparams",

device=device)

voc_predictor

voc_predictor = get_predictor(

model_dir=voc_inference_dir,

model_file="pwgan_aishell3" + ".pdmodel",  # 这里以 pwgan_aishell3 为例子，其它模型记得修改此处模型名称

params_file="pwgan_aishell3" + ".pdiparams",

device=device)

output_dir = Path(wav_output_dir)

output_dir.mkdir(parents=True, exist_ok=True)

sentences = list(text_dict.items())

merge_sentences = True

fs = 24000

for utt_id, sentence in sentences:

am_output_data = get_am_output(

  input=sentence,

  am_predictor=am_predictor,

  am="fastspeech2_mix",

  frontend=frontend,

  lang="mix",

  merge_sentences=merge_sentences,

  speaker_dict=os.path.join(am_inference_dir, "phone_id_map.txt"),

  spk_id=0, )

wav = get_voc_output(

    voc_predictor=voc_predictor, input=am_output_data)

# 保存文件

sf.write(output_dir / (utt_id + ".wav"), wav, samplerate=fs)

return

if name == 'main':

模型路径

am_inference_dir = "demo"

声码器路径，这里以 pwgan_aishell3 为例子

voc_inference_dir = "pwgan_aishell3_static_1.1.0"

音频生成的路径，修改成你音频想要保存的路径

wav_output_dir = "output"

选择设备[gpu / cpu]，这里以GPU为例子，

device = "cpu"

想要生成的文本文档对应文件名

txt_name = "文档.txt"

the_main(get_text_dict(name=am_inference_dir,txtname=txt_name)) `

报错： Traceback (most recent call last): File "E:\Record\anntest\test_2.py", line 155, in <module> the_main(get_text_dict(name=am_inference_dir,txtname=txt_name)) File "E:\Record\anntest\test_2.py", line 103, in the_main am_output_data = get_am_output( File "D:\Program Files\Python39\lib\site-packages\paddlespeech\t2s\exps\syn_utils.py", line 503, in get_am_output frontend_dict = run_frontend( File "D:\Program Files\Python39\lib\site-packages\paddlespeech\t2s\exps\syn_utils.py", line 310, in run_frontend input_ids = frontend.get_input_ids( File "D:\Program Files\Python39\lib\site-packages\paddlespeech\t2s\frontend\mix_frontend.py", line 123, in get_input_ids input_ids = self.zh_frontend.get_input_ids( File "D:\Program Files\Python39\lib\site-packages\paddlespeech\t2s\frontend\zh_frontend.py", line 540, in get_input_ids phonemes = self.get_phonemes( File "D:\Program Files\Python39\lib\site-packages\paddlespeech\t2s\frontend\zh_frontend.py", line 455, in get_phonemes phonemes = self._g2p( File "D:\Program Files\Python39\lib\site-packages\paddlespeech\t2s\frontend\zh_frontend.py", line 261, in _g2p sub_finals = self.tone_modifier.modified_tone(word, pos, File "D:\Program Files\Python39\lib\site-packages\paddlespeech\t2s\frontend\tone_sandhi.py", line 353, in modified_tone finals = self._three_sandhi(word, finals) File "D:\Program Files\Python39\lib\site-packages\paddlespeech\t2s\frontend\tone_sandhi.py", line 173, in _three_sandhi finals[0] = finals[0][:-1] + "2" IndexError: list index out of range

[iftaken]

错误看起来是语音合成前端引起的，需要你检查一下PaddleSpeech的版本以及输入的文本，看看输入是否符合要求

[ANyyJS]

谢谢回答。版本是paddlespeech1.3.0 paddlespeech-feat 0.1.0。输入的文本没有问题。

[stale[bot]]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.