Spaces:
Sleeping
Sleeping
Gong Junmin
commited on
Commit
•
260d83d
1
Parent(s):
f94ba49
add refer wav support
Browse files- app.py +30 -14
- emotion_extract.py +6 -4
app.py
CHANGED
@@ -5,6 +5,7 @@ import utils
|
|
5 |
from models import SynthesizerTrn
|
6 |
from text.symbols import symbols
|
7 |
from text import text_to_sequence
|
|
|
8 |
import numpy as np
|
9 |
|
10 |
|
@@ -32,13 +33,13 @@ emotion_dict = {
|
|
32 |
"平静2": 3554
|
33 |
}
|
34 |
import random
|
35 |
-
def tts(txt, emotion):
|
36 |
stn_tst = get_text(txt, hps)
|
37 |
randsample = None
|
38 |
with torch.no_grad():
|
39 |
x_tst = stn_tst.unsqueeze(0)
|
40 |
x_tst_lengths = torch.LongTensor([stn_tst.size(0)])
|
41 |
-
sid = torch.LongTensor([
|
42 |
if type(emotion) ==int:
|
43 |
emo = torch.FloatTensor(all_emotions[emotion]).unsqueeze(0)
|
44 |
elif emotion == "random":
|
@@ -57,54 +58,69 @@ def tts(txt, emotion):
|
|
57 |
return audio, randsample
|
58 |
|
59 |
|
60 |
-
def tts1(text, emotion):
|
61 |
if len(text) > 150:
|
62 |
return "Error: Text is too long", None
|
63 |
-
audio, _ = tts(text, emotion)
|
64 |
return "Success", (hps.data.sampling_rate, audio)
|
65 |
|
66 |
-
def tts2(text):
|
67 |
if len(text) > 150:
|
68 |
return "Error: Text is too long", None
|
69 |
-
audio, randsample = tts(text, "random_sample")
|
70 |
|
71 |
return str(randsample), (hps.data.sampling_rate, audio)
|
72 |
|
73 |
-
def tts3(text, sample):
|
74 |
if len(text) > 150:
|
75 |
return "Error: Text is too long", None
|
76 |
try:
|
77 |
-
audio, _ = tts(text, int(sample))
|
78 |
return "Success", (hps.data.sampling_rate, audio)
|
79 |
except:
|
80 |
return "输入参数不为整数或其他错误", None
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
81 |
app = gr.Blocks()
|
82 |
with app:
|
83 |
with gr.Tabs():
|
84 |
with gr.TabItem("使用预制情感合成"):
|
|
|
85 |
tts_input1 = gr.TextArea(label="日语文本", value="こんにちは。私わあやちねねです。")
|
86 |
-
tts_input2 = gr.Dropdown(label="情感", choices=list(emotion_dict.keys()),
|
87 |
tts_submit = gr.Button("合成音频", variant="primary")
|
88 |
tts_output1 = gr.Textbox(label="Message")
|
89 |
tts_output2 = gr.Audio(label="Output")
|
90 |
-
tts_submit.click(tts1, [tts_input1, tts_input2], [tts_output1, tts_output2])
|
91 |
with gr.TabItem("随机抽取训练集样本作为情感参数"):
|
|
|
92 |
tts_input1 = gr.TextArea(label="日语文本", value="こんにちは。私わあやちねねです。")
|
93 |
tts_submit = gr.Button("合成音频", variant="primary")
|
94 |
tts_output1 = gr.Textbox(label="随机样本id(可用于第三个tab中合成)")
|
95 |
tts_output2 = gr.Audio(label="Output")
|
96 |
-
tts_submit.click(tts2, [tts_input1], [tts_output1, tts_output2])
|
97 |
|
98 |
with gr.TabItem("使用情感样本id作为情感参数"):
|
99 |
-
|
100 |
tts_input1 = gr.TextArea(label="日语文本", value="こんにちは。私わあやちねねです。")
|
101 |
tts_input2 = gr.Number(label="情感样本id", value=2004)
|
102 |
tts_submit = gr.Button("合成音频", variant="primary")
|
103 |
tts_output1 = gr.Textbox(label="Message")
|
104 |
tts_output2 = gr.Audio(label="Output")
|
105 |
-
tts_submit.click(tts3, [tts_input1, tts_input2], [tts_output1, tts_output2])
|
106 |
|
107 |
with gr.TabItem("使用参考音频作为情感参数"):
|
108 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
109 |
|
110 |
app.launch()
|
|
|
5 |
from models import SynthesizerTrn
|
6 |
from text.symbols import symbols
|
7 |
from text import text_to_sequence
|
8 |
+
from emotion_extract import extract_wav
|
9 |
import numpy as np
|
10 |
|
11 |
|
|
|
33 |
"平静2": 3554
|
34 |
}
|
35 |
import random
|
36 |
+
def tts(txt, emotion, sid=0):
|
37 |
stn_tst = get_text(txt, hps)
|
38 |
randsample = None
|
39 |
with torch.no_grad():
|
40 |
x_tst = stn_tst.unsqueeze(0)
|
41 |
x_tst_lengths = torch.LongTensor([stn_tst.size(0)])
|
42 |
+
sid = torch.LongTensor([sid])
|
43 |
if type(emotion) ==int:
|
44 |
emo = torch.FloatTensor(all_emotions[emotion]).unsqueeze(0)
|
45 |
elif emotion == "random":
|
|
|
58 |
return audio, randsample
|
59 |
|
60 |
|
61 |
+
def tts1(text, emotion, sid=0):
|
62 |
if len(text) > 150:
|
63 |
return "Error: Text is too long", None
|
64 |
+
audio, _ = tts(text, emotion, sid)
|
65 |
return "Success", (hps.data.sampling_rate, audio)
|
66 |
|
67 |
+
def tts2(text, sid=0):
|
68 |
if len(text) > 150:
|
69 |
return "Error: Text is too long", None
|
70 |
+
audio, randsample = tts(text, "random_sample", sid)
|
71 |
|
72 |
return str(randsample), (hps.data.sampling_rate, audio)
|
73 |
|
74 |
+
def tts3(text, sample, sid=0):
|
75 |
if len(text) > 150:
|
76 |
return "Error: Text is too long", None
|
77 |
try:
|
78 |
+
audio, _ = tts(text, int(sample), sid)
|
79 |
return "Success", (hps.data.sampling_rate, audio)
|
80 |
except:
|
81 |
return "输入参数不为整数或其他错误", None
|
82 |
+
|
83 |
+
|
84 |
+
def tts4(refer_wav_path, text, sid=0):
|
85 |
+
audio, _ = tts(text, refer_wav_path, sid)
|
86 |
+
return "Success", (hps.data.sampling_rate, audio)
|
87 |
+
|
88 |
+
|
89 |
app = gr.Blocks()
|
90 |
with app:
|
91 |
with gr.Tabs():
|
92 |
with gr.TabItem("使用预制情感合成"):
|
93 |
+
tts_spk_id = gr.Dropdown(label="speaker", choices=list(range(hps.data.n_speakers)), value=0)
|
94 |
tts_input1 = gr.TextArea(label="日语文本", value="こんにちは。私わあやちねねです。")
|
95 |
+
tts_input2 = gr.Dropdown(label="情感", choices=list(emotion_dict.keys()), value="平静1")
|
96 |
tts_submit = gr.Button("合成音频", variant="primary")
|
97 |
tts_output1 = gr.Textbox(label="Message")
|
98 |
tts_output2 = gr.Audio(label="Output")
|
99 |
+
tts_submit.click(tts1, [tts_input1, tts_input2, tts_spk_id], [tts_output1, tts_output2])
|
100 |
with gr.TabItem("随机抽取训练集样本作为情感参数"):
|
101 |
+
tts_spk_id = gr.Dropdown(label="speaker", choices=list(range(hps.data.n_speakers)), value=0)
|
102 |
tts_input1 = gr.TextArea(label="日语文本", value="こんにちは。私わあやちねねです。")
|
103 |
tts_submit = gr.Button("合成音频", variant="primary")
|
104 |
tts_output1 = gr.Textbox(label="随机样本id(可用于第三个tab中合成)")
|
105 |
tts_output2 = gr.Audio(label="Output")
|
106 |
+
tts_submit.click(tts2, [tts_input1, tts_spk_id], [tts_output1, tts_output2])
|
107 |
|
108 |
with gr.TabItem("使用情感样本id作为情感参数"):
|
109 |
+
tts_spk_id = gr.Dropdown(label="speaker", choices=list(range(hps.data.n_speakers)), value=0)
|
110 |
tts_input1 = gr.TextArea(label="日语文本", value="こんにちは。私わあやちねねです。")
|
111 |
tts_input2 = gr.Number(label="情感样本id", value=2004)
|
112 |
tts_submit = gr.Button("合成音频", variant="primary")
|
113 |
tts_output1 = gr.Textbox(label="Message")
|
114 |
tts_output2 = gr.Audio(label="Output")
|
115 |
+
tts_submit.click(tts3, [tts_input1, tts_input2, tts_spk_id], [tts_output1, tts_output2])
|
116 |
|
117 |
with gr.TabItem("使用参考音频作为情感参数"):
|
118 |
+
tts_spk_id = gr.Dropdown(label="speaker", choices=list(range(hps.data.n_speakers)), value=0)
|
119 |
+
tts_refer_wav = gr.File(label="参考音频")
|
120 |
+
tts_input1 = gr.TextArea(label="日语文本", value="こんにちは。私わあやちねねです。")
|
121 |
+
tts_submit = gr.Button("合成音频", variant="primary")
|
122 |
+
tts_output1 = gr.Textbox(label="Message")
|
123 |
+
tts_output2 = gr.Audio(label="Output")
|
124 |
+
tts_submit.click(tts4, [tts_refer_wav, tts_input1, tts_spk_id], [tts_output1, tts_output2])
|
125 |
|
126 |
app.launch()
|
emotion_extract.py
CHANGED
@@ -74,6 +74,7 @@ def process_func(
|
|
74 |
y = processor(x, sampling_rate=sampling_rate)
|
75 |
y = y['input_values'][0]
|
76 |
y = torch.from_numpy(y).to(device)
|
|
|
77 |
|
78 |
# run through model
|
79 |
with torch.no_grad():
|
@@ -89,13 +90,13 @@ def process_func(
|
|
89 |
# wav, sr = librosa.load(f"{rootpath}/{wavname}", 16000)
|
90 |
# display(ipd.Audio(wav, rate=sr))
|
91 |
|
92 |
-
rootpath = "dataset
|
93 |
embs = []
|
94 |
wavnames = []
|
95 |
def extract_dir(path):
|
96 |
rootpath = path
|
97 |
for idx, wavname in enumerate(os.listdir(rootpath)):
|
98 |
-
wav, sr =librosa.load(f"{rootpath}/{wavname}", 16000)
|
99 |
emb = process_func(np.expand_dims(wav, 0), sr, embeddings=True)
|
100 |
embs.append(emb)
|
101 |
wavnames.append(wavname)
|
@@ -103,10 +104,11 @@ def extract_dir(path):
|
|
103 |
print(idx, wavname)
|
104 |
|
105 |
def extract_wav(path):
|
106 |
-
wav, sr = librosa.load(path, 16000)
|
107 |
emb = process_func(np.expand_dims(wav, 0), sr, embeddings=True)
|
108 |
return emb
|
109 |
|
110 |
if __name__ == '__main__':
|
111 |
-
for spk in ["serena", "koni", "nyaru","shanoa", "mana"]:
|
|
|
112 |
extract_dir(f"dataset/{spk}")
|
|
|
74 |
y = processor(x, sampling_rate=sampling_rate)
|
75 |
y = y['input_values'][0]
|
76 |
y = torch.from_numpy(y).to(device)
|
77 |
+
y = y.unsqueeze(0)
|
78 |
|
79 |
# run through model
|
80 |
with torch.no_grad():
|
|
|
90 |
# wav, sr = librosa.load(f"{rootpath}/{wavname}", 16000)
|
91 |
# display(ipd.Audio(wav, rate=sr))
|
92 |
|
93 |
+
rootpath = "dataset"
|
94 |
embs = []
|
95 |
wavnames = []
|
96 |
def extract_dir(path):
|
97 |
rootpath = path
|
98 |
for idx, wavname in enumerate(os.listdir(rootpath)):
|
99 |
+
wav, sr =librosa.load(f"{rootpath}/{wavname}", sr=16000)
|
100 |
emb = process_func(np.expand_dims(wav, 0), sr, embeddings=True)
|
101 |
embs.append(emb)
|
102 |
wavnames.append(wavname)
|
|
|
104 |
print(idx, wavname)
|
105 |
|
106 |
def extract_wav(path):
|
107 |
+
wav, sr = librosa.load(path, sr=16000)
|
108 |
emb = process_func(np.expand_dims(wav, 0), sr, embeddings=True)
|
109 |
return emb
|
110 |
|
111 |
if __name__ == '__main__':
|
112 |
+
# for spk in ["serena", "koni", "nyaru","shanoa", "mana"]:
|
113 |
+
for spk in ["dubbingx"]:
|
114 |
extract_dir(f"dataset/{spk}")
|