Spaces:
Running
on
Zero
Running
on
Zero
feat!: create app.py to initialize Gradio interface with main function
Browse files
app.py
ADDED
@@ -0,0 +1,243 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import yt_dlp
|
2 |
+
import re
|
3 |
+
import subprocess
|
4 |
+
import os
|
5 |
+
import shutil
|
6 |
+
from pydub import AudioSegment
|
7 |
+
import gradio as gr
|
8 |
+
import traceback
|
9 |
+
import logging
|
10 |
+
from inference import proc_folder_direct
|
11 |
+
from pathlib import Path
|
12 |
+
|
13 |
+
OUTPUT_FOLDER = "separation_results/"
|
14 |
+
INPUT_FOLDER = "input"
|
15 |
+
download_path = ""
|
16 |
+
|
17 |
+
def sanitize_filename(filename):
|
18 |
+
return re.sub(r'[\\/*?:"<>|]', '_', filename)
|
19 |
+
|
20 |
+
def delete_input_files(input_dir):
|
21 |
+
wav_dir = Path(input_dir) / "wav"
|
22 |
+
for wav_file in wav_dir.glob("*.wav"):
|
23 |
+
wav_file.unlink()
|
24 |
+
print(f"Deleted {wav_file}")
|
25 |
+
|
26 |
+
def download_youtube_audio_by_title(query, state=True):
|
27 |
+
if state:
|
28 |
+
delete_input_files(INPUT_FOLDER)
|
29 |
+
|
30 |
+
ydl_opts = {
|
31 |
+
'quiet': True,
|
32 |
+
'default_search': 'ytsearch',
|
33 |
+
'noplaylist': True,
|
34 |
+
'format': 'bestaudio/best',
|
35 |
+
'outtmpl': './input/wav/%(title)s.%(ext)s',
|
36 |
+
'postprocessors': [{
|
37 |
+
'key': 'FFmpegExtractAudio',
|
38 |
+
'preferredcodec': 'wav',
|
39 |
+
}],
|
40 |
+
}
|
41 |
+
|
42 |
+
with yt_dlp.YoutubeDL(ydl_opts) as ydl:
|
43 |
+
search_results = ydl.extract_info(query, download=False)
|
44 |
+
video_info = search_results['entries'][0]
|
45 |
+
video_url = video_info['webpage_url']
|
46 |
+
video_title = video_info['title']
|
47 |
+
|
48 |
+
match = re.match(r'^(.*? - .*?)(?: \[.*\]|\(.*\))?$', video_title)
|
49 |
+
formatted_title = match.group(1) if match else video_title
|
50 |
+
|
51 |
+
formatted_title = sanitize_filename(formatted_title.strip())
|
52 |
+
|
53 |
+
ydl_opts['outtmpl'] = f'./input/wav/{formatted_title}.%(ext)s'
|
54 |
+
|
55 |
+
if state:
|
56 |
+
with yt_dlp.YoutubeDL(ydl_opts) as ydl:
|
57 |
+
ydl.download([video_url])
|
58 |
+
return f'./input/wav/{formatted_title}.wav'
|
59 |
+
|
60 |
+
return formatted_title
|
61 |
+
|
62 |
+
def run_inference(model_type, config_path, start_check_point, input_dir, output_dir, device_ids="0"):
|
63 |
+
command = [
|
64 |
+
"python", "inference.py",
|
65 |
+
"--model_type", model_type,
|
66 |
+
"--config_path", config_path,
|
67 |
+
"--start_check_point", start_check_point,
|
68 |
+
"--INPUT_FOLDER", input_dir,
|
69 |
+
"--store_dir", output_dir,
|
70 |
+
"--device_ids", device_ids
|
71 |
+
]
|
72 |
+
return subprocess.run(command, check=True, capture_output=True, text=True)
|
73 |
+
|
74 |
+
def move_stems_to_parent(input_dir):
|
75 |
+
for subdir, dirs, files in os.walk(input_dir):
|
76 |
+
if subdir == input_dir:
|
77 |
+
continue
|
78 |
+
|
79 |
+
parent_dir = os.path.dirname(subdir)
|
80 |
+
song_name = os.path.basename(parent_dir)
|
81 |
+
|
82 |
+
if 'htdemucs' in subdir:
|
83 |
+
print(f"Processing htdemucs in {subdir}")
|
84 |
+
bass_path = os.path.join(subdir, f"{song_name}_bass.wav")
|
85 |
+
if os.path.exists(bass_path):
|
86 |
+
new_bass_path = os.path.join(parent_dir, "bass.wav")
|
87 |
+
print(f"Moving {bass_path} to {new_bass_path}")
|
88 |
+
shutil.move(bass_path, new_bass_path)
|
89 |
+
else:
|
90 |
+
print(f"Bass file not found: {bass_path}")
|
91 |
+
elif 'mel_band_roformer' in subdir:
|
92 |
+
print(f"Processing mel_band_roformer in {subdir}")
|
93 |
+
vocals_path = os.path.join(subdir, f"{song_name}_vocals.wav")
|
94 |
+
if os.path.exists(vocals_path):
|
95 |
+
new_vocals_path = os.path.join(parent_dir, "vocals.wav")
|
96 |
+
print(f"Moving {vocals_path} to {new_vocals_path}")
|
97 |
+
shutil.move(vocals_path, new_vocals_path)
|
98 |
+
else:
|
99 |
+
print(f"Vocals file not found: {vocals_path}")
|
100 |
+
elif 'scnet' in subdir:
|
101 |
+
print(f"Processing scnet in {subdir}")
|
102 |
+
other_path = os.path.join(subdir, f"{song_name}_other.wav")
|
103 |
+
if os.path.exists(other_path):
|
104 |
+
new_other_path = os.path.join(parent_dir, "other.wav")
|
105 |
+
print(f"Moving {other_path} to {new_other_path}")
|
106 |
+
shutil.move(other_path, new_other_path)
|
107 |
+
else:
|
108 |
+
print(f"Other file not found: {other_path}")
|
109 |
+
elif 'bs_roformer' in subdir:
|
110 |
+
print(f"Processing bs_roformer in {subdir}")
|
111 |
+
instrumental_path = os.path.join(subdir, f"{song_name}_other.wav")
|
112 |
+
if os.path.exists(instrumental_path):
|
113 |
+
new_instrumental_path = os.path.join(parent_dir, "instrumental.wav")
|
114 |
+
print(f"Moving {instrumental_path} to {new_instrumental_path}")
|
115 |
+
shutil.move(instrumental_path, new_instrumental_path)
|
116 |
+
else:
|
117 |
+
print(f"Instrumental file not found: {instrumental_path}")
|
118 |
+
|
119 |
+
def combine_stems_for_all(input_dir):
|
120 |
+
for subdir, _, _ in os.walk(input_dir):
|
121 |
+
if subdir == input_dir:
|
122 |
+
continue
|
123 |
+
|
124 |
+
song_name = os.path.basename(subdir)
|
125 |
+
print(f"Processing {subdir}")
|
126 |
+
|
127 |
+
stem_paths = {
|
128 |
+
"vocals": os.path.join(subdir, "vocals.wav"),
|
129 |
+
"bass": os.path.join(subdir, "bass.wav"),
|
130 |
+
"others": os.path.join(subdir, "other.wav"),
|
131 |
+
"instrumental": os.path.join(subdir, "instrumental.wav")
|
132 |
+
}
|
133 |
+
|
134 |
+
if not all(os.path.exists(path) for path in stem_paths.values()):
|
135 |
+
print(f"Skipping {subdir}, not all stems are present.")
|
136 |
+
continue
|
137 |
+
|
138 |
+
stems = {name: AudioSegment.from_file(path) for name, path in stem_paths.items()}
|
139 |
+
combined = stems["vocals"].overlay(stems["bass"]).overlay(stems["others"]).overlay(stems["instrumental"])
|
140 |
+
|
141 |
+
output_file = os.path.join(subdir, f"{song_name}.MDS.wav")
|
142 |
+
combined.export(output_file, format="wav")
|
143 |
+
print(f"Exported combined stems to {output_file}")
|
144 |
+
|
145 |
+
def delete_folders_and_files(input_dir):
|
146 |
+
folders_to_delete = ['htdemucs', 'mel_band_roformer', 'scnet', 'bs_roformer']
|
147 |
+
files_to_delete = ['bass.wav', 'vocals.wav', 'other.wav', 'instrumental.wav']
|
148 |
+
|
149 |
+
for root, dirs, files in os.walk(input_dir, topdown=False):
|
150 |
+
if root == input_dir:
|
151 |
+
continue
|
152 |
+
|
153 |
+
for folder in folders_to_delete:
|
154 |
+
folder_path = os.path.join(root, folder)
|
155 |
+
if os.path.isdir(folder_path):
|
156 |
+
print(f"Deleting folder: {folder_path}")
|
157 |
+
shutil.rmtree(folder_path)
|
158 |
+
|
159 |
+
for file in files_to_delete:
|
160 |
+
file_path = os.path.join(root, file)
|
161 |
+
if os.path.isfile(file_path):
|
162 |
+
print(f"Deleting file: {file_path}")
|
163 |
+
os.remove(file_path)
|
164 |
+
|
165 |
+
for root, dirs, files in os.walk(OUTPUT_FOLDER):
|
166 |
+
for dir_name in dirs:
|
167 |
+
if dir_name.endswith('_vocals'):
|
168 |
+
dir_path = os.path.join(root, dir_name)
|
169 |
+
print(f"Deleting folder: {dir_path}")
|
170 |
+
shutil.rmtree(dir_path)
|
171 |
+
|
172 |
+
print("Cleanup completed.")
|
173 |
+
|
174 |
+
def process_audio(song_title):
|
175 |
+
try:
|
176 |
+
yield "Finding audio...", None
|
177 |
+
if title_input == "":
|
178 |
+
raise ValueError("Please enter a song title.")
|
179 |
+
|
180 |
+
formatted_title = download_youtube_audio_by_title(song_title, False)
|
181 |
+
|
182 |
+
yield "Starting SCNet inference...", None
|
183 |
+
proc_folder_direct("scnet", "configs/config_scnet_other.yaml", "results/model_scnet_other.ckpt", f"{INPUT_FOLDER}/wav", OUTPUT_FOLDER)
|
184 |
+
|
185 |
+
yield "Starting Mel Band Roformer inference...", None
|
186 |
+
proc_folder_direct("mel_band_roformer", "configs/config_mel_band_roformer_vocals.yaml", "results/model_mel_band_roformer_vocals.ckpt", f"{INPUT_FOLDER}/wav", OUTPUT_FOLDER, extract_instrumental=True)
|
187 |
+
yield "Starting HTDemucs inference...", None
|
188 |
+
proc_folder_direct("htdemucs", "configs/config_htdemucs_bass.yaml", "results/model_htdemucs_bass.th", f"{INPUT_FOLDER}/wav", OUTPUT_FOLDER)
|
189 |
+
|
190 |
+
source_path = f'{OUTPUT_FOLDER}{formatted_title}/mel_band_roformer/{formatted_title}_instrumental.wav'
|
191 |
+
destination_path = f'{OUTPUT_FOLDER}{formatted_title}/mel_band_roformer/{formatted_title}.wav'
|
192 |
+
|
193 |
+
os.rename(source_path, destination_path)
|
194 |
+
|
195 |
+
yield "Starting BS Roformer inference...", None
|
196 |
+
proc_folder_direct("bs_roformer", "configs/config_bs_roformer_instrumental.yaml", "results/model_bs_roformer_instrumental.ckpt", f'{OUTPUT_FOLDER}{formatted_title}/mel_band_roformer', OUTPUT_FOLDER)
|
197 |
+
|
198 |
+
yield "Moving input files...", None
|
199 |
+
delete_input_files(INPUT_FOLDER)
|
200 |
+
|
201 |
+
yield "Moving stems to parent...", None
|
202 |
+
move_stems_to_parent(OUTPUT_FOLDER)
|
203 |
+
|
204 |
+
yield "Combining stems...", None
|
205 |
+
combine_stems_for_all(OUTPUT_FOLDER)
|
206 |
+
|
207 |
+
yield "Cleaning up...", None
|
208 |
+
delete_folders_and_files(OUTPUT_FOLDER)
|
209 |
+
|
210 |
+
|
211 |
+
yield f"Audio processing completed successfully.", f'{OUTPUT_FOLDER}{formatted_title}/{formatted_title}.MDS.wav'
|
212 |
+
except Exception as e:
|
213 |
+
error_msg = f"An error occurred: {str(e)}\n{traceback.format_exc()}"
|
214 |
+
logging.error(error_msg)
|
215 |
+
yield error_msg, None
|
216 |
+
|
217 |
+
with gr.Blocks() as demo:
|
218 |
+
gr.Markdown("# Music Player and Processor")
|
219 |
+
|
220 |
+
with gr.Row():
|
221 |
+
title_input = gr.Textbox(label="Enter Song Title")
|
222 |
+
play_button = gr.Button("Play")
|
223 |
+
|
224 |
+
audio_output = gr.Audio(label="Audio Player")
|
225 |
+
|
226 |
+
process_button = gr.Button("Process Audio")
|
227 |
+
|
228 |
+
log_output = gr.Textbox(label="Processing Log", interactive=False)
|
229 |
+
processed_audio_output = gr.Audio(label="Processed Audio")
|
230 |
+
|
231 |
+
play_button.click(
|
232 |
+
fn=download_youtube_audio_by_title,
|
233 |
+
inputs=title_input,
|
234 |
+
outputs=audio_output
|
235 |
+
)
|
236 |
+
|
237 |
+
process_button.click(
|
238 |
+
fn=process_audio,
|
239 |
+
inputs=title_input,
|
240 |
+
outputs=[log_output, processed_audio_output],
|
241 |
+
show_progress=True
|
242 |
+
)
|
243 |
+
demo.launch()
|