faver96 commited on
Commit
cdabaed
·
verified ·
1 Parent(s): 2935cde

Upload Stems.py

Browse files
Files changed (1) hide show
  1. app/pages/Stems.py +292 -0
app/pages/Stems.py ADDED
@@ -0,0 +1,292 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import os
2
+ from pathlib import Path
3
+ from typing import List
4
+ from loguru import logger as log
5
+
6
+ import requests
7
+ import streamlit as st
8
+ from streamlit_option_menu import option_menu
9
+
10
+ from footer import footer
11
+ from header import header
12
+ from helpers import (
13
+ load_audio_segment,
14
+ load_list_of_songs,
15
+ plot_audio,
16
+ url_is_valid,
17
+ file_size_is_valid,
18
+ delete_old_files,
19
+ )
20
+ from service.demucs_runner import separator
21
+ from service.vocal_remover.runner import load_model, separate
22
+ from style import CSS_TABS
23
+
24
+
25
+ label_sources = {
26
+ "no_vocals.mp3": "🎶 Instrumental",
27
+ "vocals.mp3": "🎤 Vocals",
28
+ "drums.mp3": "🥁 Drums",
29
+ "bass.mp3": "🎸 Bass",
30
+ "guitar.mp3": "🎸 Guitar",
31
+ "piano.mp3": "🎹 Piano",
32
+ "other.mp3": "🎶 Other",
33
+ }
34
+
35
+ separation_mode_to_model = {
36
+ "Vocals & Instrumental (Low Quality, Faster)": (
37
+ "vocal_remover",
38
+ ["vocals.mp3", "no_vocals.mp3"],
39
+ ),
40
+ "Vocals & Instrumental (High Quality, Slower)": ("htdemucs", ["vocals.mp3", "no_vocals.mp3"]),
41
+ "Vocals, Drums, Bass & Other (Slower)": (
42
+ "htdemucs",
43
+ ["vocals.mp3", "drums.mp3", "bass.mp3", "other.mp3"],
44
+ ),
45
+ "Vocal, Drums, Bass, Guitar, Piano & Other (Slowest)": (
46
+ "htdemucs_6s",
47
+ ["vocals.mp3", "drums.mp3", "bass.mp3", "guitar.mp3", "piano.mp3", "other.mp3"],
48
+ ),
49
+ }
50
+
51
+ extensions = ["mp3", "wav", "ogg", "flac"]
52
+
53
+ out_path = Path("/tmp")
54
+ in_path = Path("/tmp")
55
+
56
+
57
+ @st.cache_data(show_spinner=False)
58
+ def get_sources(path, file_sources):
59
+ sources = {}
60
+ for file in file_sources:
61
+ fullpath = path / file
62
+ if fullpath.exists():
63
+ sources[file] = fullpath
64
+ return sources
65
+
66
+
67
+ def reset_execution():
68
+ st.session_state.executed = False
69
+
70
+
71
+ def show_results(model_name: str, dir_name_output: str, file_sources: List):
72
+ sources = get_sources(out_path / Path(model_name) / dir_name_output, file_sources)
73
+ tab_sources = st.tabs([f"**{label_sources.get(k)}**" for k in sources.keys()])
74
+ for i, (file, pathname) in enumerate(sources.items()):
75
+ with tab_sources[i]:
76
+ cols = st.columns(2)
77
+ with cols[0]:
78
+ auseg = load_audio_segment(pathname, "mp3")
79
+ st.image(
80
+ plot_audio(
81
+ auseg,
82
+ 32767,
83
+ file=file,
84
+ model_name=model_name,
85
+ dir_name_output=dir_name_output,
86
+ ),
87
+ use_column_width="always",
88
+ )
89
+ with cols[1]:
90
+ st.audio(str(pathname))
91
+ log.info(f"Displaying results for {dir_name_output} - {model_name}")
92
+
93
+
94
+ def body():
95
+ filename = None
96
+ name_song = None
97
+ st.markdown(
98
+ "<h4><center>Extract Vocals & Instrumental from any song</center></h4>",
99
+ unsafe_allow_html=True,
100
+ )
101
+ st.markdown(CSS_TABS, unsafe_allow_html=True)
102
+
103
+ cols = st.columns([1, 4, 1, 3, 1])
104
+
105
+ with cols[1]:
106
+ with st.columns([1, 8, 1])[1]:
107
+ option = option_menu(
108
+ menu_title=None,
109
+ options=["Examples", "Upload File", "From URL"],
110
+ icons=["cloud-upload-fill", "link-45deg", "music-note-list"],
111
+ orientation="horizontal",
112
+ styles={
113
+ "container": {
114
+ "width": "100%",
115
+ "height": "3.5rem",
116
+ "margin": "0px",
117
+ "padding": "0px",
118
+ },
119
+ "icon": {"font-size": "1rem"},
120
+ "nav-link": {
121
+ "display": "flex",
122
+ "height": "3rem",
123
+ "justify-content": "center",
124
+ "align-items": "center",
125
+ "text-align": "center",
126
+ "flex-direction": "column",
127
+ "font-size": "1rem",
128
+ "padding-left": "0px",
129
+ "padding-right": "0px",
130
+ },
131
+ },
132
+ key="option_separate",
133
+ )
134
+ if option == "Examples":
135
+ samples_song = load_list_of_songs(path="separate_songs.json")
136
+ if samples_song is not None:
137
+ name_song = st.selectbox(
138
+ label="Select a sample song and listen to sources separated",
139
+ options=list(samples_song.keys()) + [""],
140
+ format_func=lambda x: x.replace("_", " "),
141
+ index=len(samples_song),
142
+ key="select_example",
143
+ )
144
+ full_path = f"{in_path}/{name_song}"
145
+ if name_song != "" and Path(full_path).exists():
146
+ st.audio(full_path)
147
+ else:
148
+ name_song = None
149
+
150
+ elif option == "Upload File":
151
+ uploaded_file = st.file_uploader(
152
+ "Choose a file",
153
+ type=extensions,
154
+ key="file",
155
+ help="Supported formats: mp3, wav, ogg, flac.",
156
+ )
157
+ if uploaded_file is not None:
158
+ with st.spinner("Loading audio..."):
159
+ with open(in_path / uploaded_file.name, "wb") as f:
160
+ f.write(uploaded_file.getbuffer())
161
+ filename = uploaded_file.name
162
+ st.audio(uploaded_file)
163
+
164
+ elif option == "From URL":
165
+ url = st.text_input(
166
+ "Paste the URL of the audio file",
167
+ key="url_input",
168
+ help="Supported formats: mp3, wav, ogg, flac.",
169
+ )
170
+
171
+ if url != "" and url_is_valid(url):
172
+ with st.spinner("Downloading audio..."):
173
+ filename = url.split("/")[-1]
174
+ response = requests.get(url, stream=True)
175
+ if response.status_code == 200 and file_size_is_valid(
176
+ response.headers.get("Content-Length")
177
+ ):
178
+ file_size = 0
179
+ with open(in_path / filename, "wb") as audio_file:
180
+ for chunk in response.iter_content(chunk_size=1024):
181
+ if chunk:
182
+ audio_file.write(chunk)
183
+ file_size += len(chunk)
184
+ if not file_size_is_valid(file_size):
185
+ audio_file.close()
186
+ os.remove(in_path / filename)
187
+ filename = None
188
+ return
189
+ st.audio(f"{in_path}/{filename}")
190
+ else:
191
+ st.error(
192
+ "Failed to download audio file. Try to download it manually and upload it."
193
+ )
194
+ filename = None
195
+
196
+ with cols[3]:
197
+ separation_mode = st.selectbox(
198
+ "Choose the separation mode",
199
+ [
200
+ "Vocals & Instrumental (Low Quality, Faster)",
201
+ "Vocals & Instrumental (High Quality, Slower)",
202
+ "Vocals, Drums, Bass & Other (Slower)",
203
+ "Vocal, Drums, Bass, Guitar, Piano & Other (Slowest)",
204
+ ],
205
+ on_change=reset_execution(),
206
+ key="separation_mode",
207
+ )
208
+ if separation_mode == "Vocals & Instrumental (Low Quality, Faster)":
209
+ max_duration = 30
210
+ else:
211
+ max_duration = 15
212
+ model_name, file_sources = separation_mode_to_model[separation_mode]
213
+
214
+ if filename is not None:
215
+ song = load_audio_segment(in_path / filename, filename.split(".")[-1])
216
+ n_secs = round(len(song) / 1000)
217
+ if os.environ.get("ENV_LIMITATION", False):
218
+ with cols[3]:
219
+ start_time = st.number_input(
220
+ "Choose the start time",
221
+ min_value=0,
222
+ max_value=n_secs,
223
+ step=1,
224
+ value=0,
225
+ help=f"Maximum duration is {max_duration} seconds for this separation mode.\nDuplicate this space to [remove any limit](https://github.com/fabiogra/moseca#are-there-any-limitations).",
226
+ format="%d",
227
+ )
228
+ st.session_state.start_time = start_time
229
+ end_time = min(start_time + max_duration, n_secs)
230
+ song = song[start_time * 1000 : end_time * 1000]
231
+ st.info(
232
+ f"Audio source will be processed from {start_time} to {end_time} seconds.\nDuplicate this space to [remove any limit](https://github.com/fabiogra/moseca#are-there-any-limitations).",
233
+ icon="⏱",
234
+ )
235
+ else:
236
+ start_time = 0
237
+ end_time = n_secs
238
+ with st.columns([2, 1, 2])[1]:
239
+ execute = st.button(
240
+ "Separate Music Sources 🎶", type="primary", use_container_width=True
241
+ )
242
+ if execute or st.session_state.executed:
243
+ if execute:
244
+ st.session_state.executed = False
245
+
246
+ if not st.session_state.executed:
247
+ log.info(f"{option} - Separating {filename} with {separation_mode}...")
248
+ song.export(in_path / filename, format=filename.split(".")[-1])
249
+ with st.columns([1, 1, 1])[1]:
250
+ with st.spinner("Separating source audio, it will take a while..."):
251
+ if model_name == "vocal_remover":
252
+ model, device = load_model(pretrained_model="baseline.pth")
253
+ separate(
254
+ input=in_path / filename,
255
+ model=model,
256
+ device=device,
257
+ output_dir=out_path,
258
+ )
259
+ else:
260
+ stem = None
261
+ if separation_mode == "Vocals & Instrumental (High Quality, Slower)":
262
+ stem = "vocals"
263
+
264
+ separator(
265
+ tracks=[in_path / filename],
266
+ out=out_path,
267
+ model=model_name,
268
+ shifts=1,
269
+ overlap=0.5,
270
+ stem=stem,
271
+ int24=False,
272
+ float32=False,
273
+ clip_mode="rescale",
274
+ mp3=True,
275
+ mp3_bitrate=320,
276
+ verbose=True,
277
+ start_time=start_time,
278
+ end_time=end_time,
279
+ )
280
+ dir_name_output = ".".join(filename.split(".")[:-1])
281
+ filename = None
282
+ st.session_state.executed = True
283
+ show_results(model_name, dir_name_output, file_sources)
284
+ elif name_song is not None and option == "Examples":
285
+ show_results(model_name, name_song, file_sources)
286
+
287
+
288
+ if __name__ == "__main__":
289
+ header()
290
+ body()
291
+ footer()
292
+ delete_old_files("/tmp", 60 * 30)