Spaces:
Build error
Build error
init
Browse files- app.py +4 -3
- infer/__init__.py +1 -0
app.py
CHANGED
@@ -41,12 +41,13 @@ def vc_fn(speaker, ds, vc_transform):
|
|
41 |
ph_dur= [float(i) for i in ph_dur.split(" ")]
|
42 |
f0_seq = [float(i) for i in f0_seq.split(" ")]
|
43 |
dur+=sum(ph_dur)
|
44 |
-
|
|
|
45 |
flag = True
|
46 |
except:
|
47 |
return "ds工程需要冻结f0和音素参数才能使用此模型合成", None
|
48 |
if flag:
|
49 |
-
return "单个切片时长必须小于
|
50 |
if dur>120:
|
51 |
return "总时长需要小于2分钟,否则请使用本地推理", None
|
52 |
out_audio = infer_ds(net_g, hps, ds, speaker, vc_transform)
|
@@ -69,7 +70,7 @@ with app:
|
|
69 |
|
70 |
由于此模型是nomidi模型,因此导出ds工程时需要冻结音素和音高参数, 否则会报错,具体DiffSinger工程制作详细问题可以加入DiffSinger QQ交流群 907879266
|
71 |
|
72 |
-
在线推理限制为总时长小于2分钟,且单个切片时长小于
|
73 |
""")
|
74 |
sid = gr.Dropdown(label="音色", choices=["taffy", "otto"], value="taffy")
|
75 |
vc_input3 = gr.TextArea(label="ds工程(json格式)",value='''[
|
|
|
41 |
ph_dur= [float(i) for i in ph_dur.split(" ")]
|
42 |
f0_seq = [float(i) for i in f0_seq.split(" ")]
|
43 |
dur+=sum(ph_dur)
|
44 |
+
print(sum(ph_dur))
|
45 |
+
if sum(ph_dur) >20:
|
46 |
flag = True
|
47 |
except:
|
48 |
return "ds工程需要冻结f0和音素参数才能使用此模型合成", None
|
49 |
if flag:
|
50 |
+
return "单个切片时长必须小于20s,否则请使用本地推理", None
|
51 |
if dur>120:
|
52 |
return "总时长需要小于2分钟,否则请使用本地推理", None
|
53 |
out_audio = infer_ds(net_g, hps, ds, speaker, vc_transform)
|
|
|
70 |
|
71 |
由于此模型是nomidi模型,因此导出ds工程时需要冻结音素和音高参数, 否则会报错,具体DiffSinger工程制作详细问题可以加入DiffSinger QQ交流群 907879266
|
72 |
|
73 |
+
在线推理限制为总时长小于2分钟,且单个切片时长小于20s,有更大需求请下载本仓库或github仓库代码运行ds_inference.py进行本地推理
|
74 |
""")
|
75 |
sid = gr.Dropdown(label="音色", choices=["taffy", "otto"], value="taffy")
|
76 |
vc_input3 = gr.TextArea(label="ds工程(json格式)",value='''[
|
infer/__init__.py
CHANGED
@@ -112,6 +112,7 @@ def infer_ds(model, hps, ds, speaker, trans):
|
|
112 |
result = cross_fade(result, seg_audio, current_length + silent_length)
|
113 |
current_length = current_length + silent_length + seg_audio.shape[0]
|
114 |
print("infer time:", time.time() - t1)
|
|
|
115 |
return result
|
116 |
|
117 |
|
|
|
112 |
result = cross_fade(result, seg_audio, current_length + silent_length)
|
113 |
current_length = current_length + silent_length + seg_audio.shape[0]
|
114 |
print("infer time:", time.time() - t1)
|
115 |
+
|
116 |
return result
|
117 |
|
118 |
|