Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -15,10 +15,10 @@ import pytz
|
|
15 |
|
16 |
net_g = None
|
17 |
models = {
|
18 |
-
"
|
19 |
"MistyNikki": "./MODELS/nikki9400.pth",
|
20 |
"Silverleg": "./MODELS/J8900.pth",
|
21 |
-
"
|
22 |
"LucidMoon": "./MODELS/lucid.pth",
|
23 |
"Rrabbitt": "./MODELS/rabbit4900.pth",
|
24 |
"VivaciousViolet": "./MODELS/vv.pth",
|
@@ -71,10 +71,14 @@ def infer(text, sdp_ratio, noise_scale, noise_scale_w, length_scale, sid, model_
|
|
71 |
sf.write("tmp.wav", audio, 44100)
|
72 |
return audio
|
73 |
|
74 |
-
def
|
75 |
tz = pytz.timezone('Asia/Shanghai')
|
76 |
now = datetime.now(tz).strftime('%m%d%H%M%S')
|
|
|
|
|
|
|
77 |
os.makedirs('out', exist_ok=True)
|
|
|
78 |
output_path_mp3 = os.path.join('out', f"{now}.mp3")
|
79 |
|
80 |
renamed_input_path = os.path.join('in', f"in.wav")
|
@@ -88,7 +92,7 @@ def tts_generator(text, sdp_ratio, noise_scale, noise_scale_w, length_scale, mod
|
|
88 |
global net_g,speakers
|
89 |
model_path = models[model]
|
90 |
net_g, _, _, _ = utils.load_checkpoint(model_path, net_g, None, skip_optimizer=True)
|
91 |
-
print(text)
|
92 |
try:
|
93 |
with torch.no_grad():
|
94 |
audio = infer(text, sdp_ratio=sdp_ratio, noise_scale=noise_scale, noise_scale_w=noise_scale_w, length_scale=length_scale, sid=speaker,model_dir=model)
|
@@ -121,12 +125,12 @@ if __name__ == "__main__":
|
|
121 |
|
122 |
gr.Markdown('''
|
123 |
**测试用**\n
|
124 |
-
|
125 |
-
huggingface.co/spaces/Ailyth/Multi-voice-TTS/tree/main
|
126 |
''')
|
127 |
text = gr.TextArea(label="输入需要生成语音的文字", placeholder="输入文字",
|
128 |
value="在不在?能不能借给我三百块钱买可乐",
|
129 |
-
info="使用huggingface的免费CPU进行推理,因此速度不快,一次性不要输入超过500
|
130 |
)
|
131 |
model = gr.Radio(choices=list(models.keys()), value=list(models.keys())[0], label='音声模型')
|
132 |
with gr.Accordion(label="展开设置生成参数", open=False):
|
|
|
15 |
|
16 |
net_g = None
|
17 |
models = {
|
18 |
+
"Mellowdear": "./MODELS/adorabledarling.pth",
|
19 |
"MistyNikki": "./MODELS/nikki9400.pth",
|
20 |
"Silverleg": "./MODELS/J8900.pth",
|
21 |
+
"Umemura": "./MODELS/take2.pth",
|
22 |
"LucidMoon": "./MODELS/lucid.pth",
|
23 |
"Rrabbitt": "./MODELS/rabbit4900.pth",
|
24 |
"VivaciousViolet": "./MODELS/vv.pth",
|
|
|
71 |
sf.write("tmp.wav", audio, 44100)
|
72 |
return audio
|
73 |
|
74 |
+
def now():
|
75 |
tz = pytz.timezone('Asia/Shanghai')
|
76 |
now = datetime.now(tz).strftime('%m%d%H%M%S')
|
77 |
+
return now
|
78 |
+
|
79 |
+
def convert_wav_to_mp3(wav_file):
|
80 |
os.makedirs('out', exist_ok=True)
|
81 |
+
now=now()
|
82 |
output_path_mp3 = os.path.join('out', f"{now}.mp3")
|
83 |
|
84 |
renamed_input_path = os.path.join('in', f"in.wav")
|
|
|
92 |
global net_g,speakers
|
93 |
model_path = models[model]
|
94 |
net_g, _, _, _ = utils.load_checkpoint(model_path, net_g, None, skip_optimizer=True)
|
95 |
+
print(now()+text)
|
96 |
try:
|
97 |
with torch.no_grad():
|
98 |
audio = infer(text, sdp_ratio=sdp_ratio, noise_scale=noise_scale, noise_scale_w=noise_scale_w, length_scale=length_scale, sid=speaker,model_dir=model)
|
|
|
125 |
|
126 |
gr.Markdown('''
|
127 |
**测试用**\n
|
128 |
+
项目文件:\n
|
129 |
+
huggingface.co/spaces/Ailyth/Multi-voice-TTS/tree/main
|
130 |
''')
|
131 |
text = gr.TextArea(label="输入需要生成语音的文字", placeholder="输入文字",
|
132 |
value="在不在?能不能借给我三百块钱买可乐",
|
133 |
+
info="使用huggingface的免费CPU进行推理,因此速度不快,一次性不要输入超过500汉字。字数越多,生成速度越慢,请耐心等待,只会说中文。",
|
134 |
)
|
135 |
model = gr.Radio(choices=list(models.keys()), value=list(models.keys())[0], label='音声模型')
|
136 |
with gr.Accordion(label="展开设置生成参数", open=False):
|