Mahiruoshi commited on
Commit
e822849
1 Parent(s): 9060ca6

Update app.py

Browse files
Files changed (1) hide show
  1. app.py +4 -37
app.py CHANGED
@@ -1,35 +1,3 @@
1
- import ONNXVITS_models
2
- import utils
3
- from text import text_to_sequence
4
- import torch
5
- import commons
6
-
7
- def get_text(text, hps):
8
- text_norm = text_to_sequence(text, hps.symbols, hps.data.text_cleaners)
9
- if hps.data.add_blank:
10
- text_norm = commons.intersperse(text_norm, 0)
11
- text_norm = torch.LongTensor(text_norm)
12
- return text_norm
13
-
14
- hps = utils.get_hparams_from_file("lovelive/config.json")
15
- symbols = hps.symbols
16
- net_g = ONNXVITS_models.SynthesizerTrn(
17
- len(symbols),
18
- hps.data.filter_length // 2 + 1,
19
- hps.train.segment_size // hps.data.hop_length,
20
- n_speakers=hps.data.n_speakers,
21
- **hps.model)
22
- _ = net_g.eval()
23
- _ = utils.load_checkpoint("lovelive/G_525000.pth", net_g)
24
-
25
- text1 = get_text("[JA]ありがとうございます。[JA]", hps)
26
- stn_tst = text1
27
- with torch.no_grad():
28
- x_tst = stn_tst.unsqueeze(0)
29
- x_tst_lengths = torch.LongTensor([stn_tst.size(0)])
30
- sid = torch.tensor([0])
31
- o = net_g(x_tst, x_tst_lengths, sid=sid, noise_scale=.667, noise_scale_w=0.8, length_scale=1)
32
- '''
33
  import romajitable
34
  import re
35
  import numpy as np
@@ -39,7 +7,7 @@ import IPython.display as ipd
39
  import torch
40
  import commons
41
  import utils
42
- import ONNXVITS_infer
43
  from text.symbols import symbols
44
  from text import text_to_sequence
45
  import gradio as gr
@@ -181,7 +149,7 @@ lan = ["中文","日文","自动"]
181
  idols = ["高咲侑","歩夢","かすみ","しずく","果林","愛","せつ菜","璃奈","栞子","エマ","ランジュ","ミア","派蒙"]
182
  dev = torch.device("cpu")
183
  hps_ms = utils.get_hparams_from_file("lovelive/config.json")
184
- net_g_ms = ONNXVITS_infer.SynthesizerTrn(
185
  len(symbols),
186
  hps_ms.data.filter_length // 2 + 1,
187
  hps_ms.train.segment_size // hps_ms.data.hop_length,
@@ -193,7 +161,7 @@ app = gr.Blocks()
193
  with app:
194
  with gr.Tabs():
195
 
196
- with gr.TabItem("虹团vits模型,现可按句分割实现长文本合成,onnx导出后存在质量损失,建议本地运行vits模型"):
197
 
198
  tts_input1 = gr.TextArea(label="去标贝新模型,老版本在lovelive文件夹中", value="数千怀言者已经为你集结,列队在通往主舰桥的过道上。他们歌唱着你们名字,高声呼喊,以一种原始的、咆哮般的合唱作为对你的致敬。你从他们中间走过,一边点头,一边接受他们的赞美,你沉溺其中,几乎被他们巨大的音量所震撼。\n他们之中没有一个胆敢直视你。没有一个能够承受。你对他们超人类的眼睛来说都太过光辉。从他们中间走过时,你巨大的影子从他们身上掠过,他们立时将目光挪开,眼含泪水,吟诵你的大名时甚至不敢看你一眼。他们的吟唱中含有愤怒。几乎是疯狂的绝望。那感觉就好像他们害怕停下来,害怕自己会喘息停顿,好像尖叫出你的名字是唯一能让他们活着的事情。\n或许确实如此。作为对他们崇拜的回应,你谦虚地抬抬手,随后走进主舰桥。\nI In a word, Horus is a joker.")
199
  language = gr.Dropdown(label="选择语言,目前勉强可以做到自动识别",choices=lan, value="自动", interactive=True)
@@ -205,5 +173,4 @@ with app:
205
  tts_output2 = gr.Audio(label="Output")
206
  tts_submit.click(infer, [language,tts_input1,speaker1,para_input1,para_input2,para_input3], [tts_output2])
207
  #app.launch(share=True)
208
- app.launch()
209
- '''
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  import romajitable
2
  import re
3
  import numpy as np
 
7
  import torch
8
  import commons
9
  import utils
10
+ from models import SynthesizerTrn
11
  from text.symbols import symbols
12
  from text import text_to_sequence
13
  import gradio as gr
 
149
  idols = ["高咲侑","歩夢","かすみ","しずく","果林","愛","せつ菜","璃奈","栞子","エマ","ランジュ","ミア","派蒙"]
150
  dev = torch.device("cpu")
151
  hps_ms = utils.get_hparams_from_file("lovelive/config.json")
152
+ net_g_ms = SynthesizerTrn(
153
  len(symbols),
154
  hps_ms.data.filter_length // 2 + 1,
155
  hps_ms.train.segment_size // hps_ms.data.hop_length,
 
161
  with app:
162
  with gr.Tabs():
163
 
164
+ with gr.TabItem("虹团vits模型,现可按句分割实现长文本合成,可自行用export_to_onnx.py导出"):
165
 
166
  tts_input1 = gr.TextArea(label="去标贝新模型,老版本在lovelive文件夹中", value="数千怀言者已经为你集结,列队在通往主舰桥的过道上。他们歌唱着你们名字,高声呼喊,以一种原始的、咆哮般的合唱作为对你的致敬。你从他们中间走过,一边点头,一边接受他们的赞美,你沉溺其中,几乎被他们巨大的音量所震撼。\n他们之中没有一个胆敢直视你。没有一个能够承受。你对他们超人类的眼睛来说都太过光辉。从他们中间走过时,你巨大的影子从他们身上掠过,他们立时将目光挪开,眼含泪水,吟诵你的大名时甚至不敢看你一眼。他们的吟唱中含有愤怒。几乎是疯狂的绝望。那感觉就好像他们害怕停下来,害怕自己会喘息停顿,好像尖叫出你的名字是唯一能让他们活着的事情。\n或许确实如此。作为对他们崇拜的回应,你谦虚地抬抬手,随后走进主舰桥。\nI In a word, Horus is a joker.")
167
  language = gr.Dropdown(label="选择语言,目前勉强可以做到自动识别",choices=lan, value="自动", interactive=True)
 
173
  tts_output2 = gr.Audio(label="Output")
174
  tts_submit.click(infer, [language,tts_input1,speaker1,para_input1,para_input2,para_input3], [tts_output2])
175
  #app.launch(share=True)
176
+ app.launch()