Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,194 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from original import *
|
2 |
+
import shutil, glob
|
3 |
+
from easyfuncs import download_from_url, CachedModels
|
4 |
+
os.makedirs("dataset",exist_ok=True)
|
5 |
+
model_library = CachedModels()
|
6 |
+
|
7 |
+
with gr.Blocks(title="🔊",theme=gr.themes.Base(primary_hue="rose",neutral_hue="zinc")) as app:
|
8 |
+
with gr.Row():
|
9 |
+
gr.HTML("<img src='file/a.png' alt='image'>")
|
10 |
+
with gr.Tabs():
|
11 |
+
with gr.TabItem("Inference"):
|
12 |
+
with gr.Row():
|
13 |
+
voice_model = gr.Dropdown(label="Model Voice", choices=sorted(names), value=lambda:sorted(names)[0] if len(sorted(names)) > 0 else '', interactive=True)
|
14 |
+
refresh_button = gr.Button("Refresh", variant="primary")
|
15 |
+
spk_item = gr.Slider(
|
16 |
+
minimum=0,
|
17 |
+
maximum=2333,
|
18 |
+
step=1,
|
19 |
+
label="Speaker ID",
|
20 |
+
value=0,
|
21 |
+
visible=False,
|
22 |
+
interactive=True,
|
23 |
+
)
|
24 |
+
vc_transform0 = gr.Number(
|
25 |
+
label="Pitch",
|
26 |
+
value=0
|
27 |
+
)
|
28 |
+
but0 = gr.Button(value="Convert", variant="primary")
|
29 |
+
with gr.Row():
|
30 |
+
with gr.Column():
|
31 |
+
with gr.Row():
|
32 |
+
dropbox = gr.File(label="Drop your audio here & hit the Reload button.")
|
33 |
+
with gr.Row():
|
34 |
+
record_button=gr.Audio(source="microphone", label="OR Record audio.", type="filepath")
|
35 |
+
with gr.Row():
|
36 |
+
paths_for_files = lambda path:[os.path.abspath(os.path.join(path, f)) for f in os.listdir(path) if os.path.splitext(f)[1].lower() in ('.mp3', '.wav', '.flac', '.ogg')]
|
37 |
+
input_audio0 = gr.Dropdown(
|
38 |
+
label="Input Path",
|
39 |
+
value=paths_for_files('audios')[0] if len(paths_for_files('audios')) > 0 else '',
|
40 |
+
choices=paths_for_files('audios'), # Only show absolute paths for audio files ending in .mp3, .wav, .flac or .ogg
|
41 |
+
allow_custom_value=True
|
42 |
+
)
|
43 |
+
with gr.Row():
|
44 |
+
audio_player = gr.Audio()
|
45 |
+
input_audio0.change(
|
46 |
+
inputs=[input_audio0],
|
47 |
+
outputs=[audio_player],
|
48 |
+
fn=lambda path: {"value":path,"__type__":"update"} if os.path.exists(path) else None
|
49 |
+
)
|
50 |
+
record_button.stop_recording(
|
51 |
+
fn=lambda audio:audio, #TODO save wav lambda
|
52 |
+
inputs=[record_button],
|
53 |
+
outputs=[input_audio0])
|
54 |
+
dropbox.upload(
|
55 |
+
fn=lambda audio:audio.name,
|
56 |
+
inputs=[dropbox],
|
57 |
+
outputs=[input_audio0])
|
58 |
+
with gr.Column():
|
59 |
+
with gr.Accordion("Change Index", open=False):
|
60 |
+
file_index2 = gr.Dropdown(
|
61 |
+
label="Change Index",
|
62 |
+
choices=sorted(index_paths),
|
63 |
+
interactive=True,
|
64 |
+
value=sorted(index_paths)[0] if len(sorted(index_paths)) > 0 else ''
|
65 |
+
)
|
66 |
+
index_rate1 = gr.Slider(
|
67 |
+
minimum=0,
|
68 |
+
maximum=1,
|
69 |
+
label="Index Strength",
|
70 |
+
value=0.5,
|
71 |
+
interactive=True,
|
72 |
+
)
|
73 |
+
vc_output2 = gr.Audio(label="Output")
|
74 |
+
with gr.Accordion("General Settings", open=False):
|
75 |
+
f0method0 = gr.Radio(
|
76 |
+
label="Method",
|
77 |
+
choices=["pm", "harvest", "crepe", "rmvpe"]
|
78 |
+
if config.dml == False
|
79 |
+
else ["pm", "harvest", "rmvpe"],
|
80 |
+
value="rmvpe",
|
81 |
+
interactive=True,
|
82 |
+
)
|
83 |
+
filter_radius0 = gr.Slider(
|
84 |
+
minimum=0,
|
85 |
+
maximum=7,
|
86 |
+
label="Breathiness Reduction (Harvest only)",
|
87 |
+
value=3,
|
88 |
+
step=1,
|
89 |
+
interactive=True,
|
90 |
+
)
|
91 |
+
resample_sr0 = gr.Slider(
|
92 |
+
minimum=0,
|
93 |
+
maximum=48000,
|
94 |
+
label="Resample",
|
95 |
+
value=0,
|
96 |
+
step=1,
|
97 |
+
interactive=True,
|
98 |
+
visible=False
|
99 |
+
)
|
100 |
+
rms_mix_rate0 = gr.Slider(
|
101 |
+
minimum=0,
|
102 |
+
maximum=1,
|
103 |
+
label="Volume Normalization",
|
104 |
+
value=0,
|
105 |
+
interactive=True,
|
106 |
+
)
|
107 |
+
protect0 = gr.Slider(
|
108 |
+
minimum=0,
|
109 |
+
maximum=0.5,
|
110 |
+
label="Breathiness Protection (0 is enabled, 0.5 is disabled)",
|
111 |
+
value=0.33,
|
112 |
+
step=0.01,
|
113 |
+
interactive=True,
|
114 |
+
)
|
115 |
+
if voice_model != None: vc.get_vc(voice_model.value,protect0,protect0)
|
116 |
+
file_index1 = gr.Textbox(
|
117 |
+
label="Index Path",
|
118 |
+
interactive=True,
|
119 |
+
visible=False#Not used here
|
120 |
+
)
|
121 |
+
refresh_button.click(
|
122 |
+
fn=change_choices,
|
123 |
+
inputs=[],
|
124 |
+
outputs=[voice_model, file_index2],
|
125 |
+
api_name="infer_refresh",
|
126 |
+
)
|
127 |
+
refresh_button.click(
|
128 |
+
fn=lambda:{"choices":paths_for_files('audios'),"__type__":"update"}, #TODO check if properly returns a sorted list of audio files in the 'audios' folder that have the extensions '.wav', '.mp3', '.ogg', or '.flac'
|
129 |
+
inputs=[],
|
130 |
+
outputs = [input_audio0],
|
131 |
+
)
|
132 |
+
refresh_button.click(
|
133 |
+
fn=lambda:{"value":paths_for_files('audios')[0],"__type__":"update"} if len(paths_for_files('audios')) > 0 else {"value":"","__type__":"update"}, #TODO check if properly returns a sorted list of audio files in the 'audios' folder that have the extensions '.wav', '.mp3', '.ogg', or '.flac'
|
134 |
+
inputs=[],
|
135 |
+
outputs = [input_audio0],
|
136 |
+
)
|
137 |
+
with gr.Row():
|
138 |
+
f0_file = gr.File(label="F0 Path", visible=False)
|
139 |
+
with gr.Row():
|
140 |
+
vc_output1 = gr.Textbox(label="Information", placeholder="Welcome!",visible=False)
|
141 |
+
but0.click(
|
142 |
+
vc.vc_single,
|
143 |
+
[
|
144 |
+
spk_item,
|
145 |
+
input_audio0,
|
146 |
+
vc_transform0,
|
147 |
+
f0_file,
|
148 |
+
f0method0,
|
149 |
+
file_index1,
|
150 |
+
file_index2,
|
151 |
+
index_rate1,
|
152 |
+
filter_radius0,
|
153 |
+
resample_sr0,
|
154 |
+
rms_mix_rate0,
|
155 |
+
protect0,
|
156 |
+
],
|
157 |
+
[vc_output1, vc_output2],
|
158 |
+
api_name="infer_convert",
|
159 |
+
)
|
160 |
+
voice_model.change(
|
161 |
+
fn=vc.get_vc,
|
162 |
+
inputs=[voice_model, protect0, protect0],
|
163 |
+
outputs=[spk_item, protect0, protect0, file_index2, file_index2],
|
164 |
+
api_name="infer_change_voice",
|
165 |
+
)
|
166 |
+
with gr.TabItem("Download Models"):
|
167 |
+
with gr.Row():
|
168 |
+
url_input = gr.Textbox(label="URL to model", value="",placeholder="https://...", scale=6)
|
169 |
+
name_output = gr.Textbox(label="Save as", value="",placeholder="MyModel",scale=2)
|
170 |
+
url_download = gr.Button(value="Download Model",scale=2)
|
171 |
+
url_download.click(
|
172 |
+
inputs=[url_input,name_output],
|
173 |
+
outputs=[url_input],
|
174 |
+
fn=download_from_url,
|
175 |
+
)
|
176 |
+
with gr.Row():
|
177 |
+
model_browser = gr.Dropdown(choices=list(model_library.models.keys()),label="OR Search Models (Quality UNKNOWN)",scale=5)
|
178 |
+
download_from_browser = gr.Button(value="Get",scale=2)
|
179 |
+
download_from_browser.click(
|
180 |
+
inputs=[model_browser],
|
181 |
+
outputs=[model_browser],
|
182 |
+
fn=lambda model: download_from_url(model_library.models[model],model),
|
183 |
+
)
|
184 |
+
|
185 |
+
|
186 |
+
if config.iscolab:
|
187 |
+
app.queue(concurrency_count=511, max_size=1022).launch(share=True)
|
188 |
+
else:
|
189 |
+
app.queue(concurrency_count=511, max_size=1022).launch(
|
190 |
+
server_name="0.0.0.0",
|
191 |
+
inbrowser=not config.noautoopen,
|
192 |
+
server_port=config.listen_port,
|
193 |
+
quiet=True,
|
194 |
+
)
|