ASesYusuf1 commited on
Commit
07f3aee
·
verified ·
1 Parent(s): 0b49175

Update gui.py

Browse files
Files changed (1) hide show
  1. gui.py +58 -916
gui.py CHANGED
@@ -1,958 +1,100 @@
1
  import gradio as gr
2
  import os
3
- import glob
4
- import subprocess
5
- from pathlib import Path
6
- from datetime import datetime
7
  import json
8
- import sys
9
- import time
10
- import random
11
- from helpers import update_model_dropdown, handle_file_upload, clear_old_output, save_uploaded_file, update_file_list, clean_model
12
- from download import download_callback
13
- from model import get_model_config, MODEL_CONFIGS
14
- from processing import process_audio, auto_ensemble_process, ensemble_audio_fn, refresh_auto_output
15
- from assets.i18n.i18n import I18nAuto
16
- from config_manager import load_config, save_config, update_favorites, save_preset, delete_preset
17
  import logging
 
 
 
 
 
18
 
19
  logging.basicConfig(filename='sesa_gui.log', level=logging.DEBUG)
20
 
21
- # BASE_DIR definition
22
  BASE_DIR = os.path.dirname(os.path.abspath(__file__))
23
  CONFIG_DIR = os.path.join(BASE_DIR, "assets")
24
  CONFIG_FILE = os.path.join(CONFIG_DIR, "config.json")
25
- URL_FILE = os.path.join(CONFIG_DIR, "last_url.txt")
26
 
27
- # Load user config at startup
28
  user_config = load_config()
29
  initial_settings = user_config["settings"]
30
  initial_favorites = user_config["favorites"]
31
- initial_presets = user_config["presets"]
32
-
33
- # Ensure auto_category is valid
34
- if "auto_category" not in initial_settings or initial_settings["auto_category"] not in MODEL_CONFIGS:
35
- initial_settings["auto_category"] = "Vocal Models"
36
-
37
- # Create config file if it doesn't exist
38
- if not os.path.exists(CONFIG_FILE):
39
- default_config = {
40
- "lang": {"override": False, "selected_lang": "auto"},
41
- "sharing": {
42
- "method": "gradio",
43
- "ngrok_token": "",
44
- "port": random.randint(1000, 9000)
45
- }
46
- }
47
- os.makedirs(CONFIG_DIR, exist_ok=True)
48
- with open(CONFIG_FILE, "w", encoding="utf-8") as f:
49
- json.dump(default_config, f, indent=2)
50
- else:
51
- try:
52
- with open(CONFIG_FILE, "r", encoding="utf-8") as f:
53
- config = json.load(f)
54
- if "lang" not in config:
55
- config["lang"] = {"override": False, "selected_lang": "auto"}
56
- if "sharing" not in config:
57
- config["sharing"] = {
58
- "method": "gradio",
59
- "ngrok_token": "",
60
- "port": random.randint(1000, 9000)
61
- }
62
- with open(CONFIG_FILE, "w", encoding="utf-8") as f:
63
- json.dump(config, f, indent=2)
64
- except json.JSONDecodeError:
65
- print("Warning: config.json is corrupted. Creating a new one.")
66
- default_config = {
67
- "lang": {"override": False, "selected_lang": "auto"},
68
- "sharing": {
69
- "method": "gradio",
70
- "ngrok_token": "",
71
- "port": random.randint(1000, 9000)
72
- }
73
- }
74
- with open(CONFIG_FILE, "w", encoding="utf-8") as f:
75
- json.dump(default_config, f, indent=2)
76
 
77
  # Initialize I18nAuto
78
  i18n = I18nAuto()
79
 
80
- # Output formats
81
- OUTPUT_FORMATS = ['wav', 'flac', 'mp3', 'ogg', 'opus', 'm4a', 'aiff', 'ac3']
82
-
83
- # Interface creation function
84
  def create_interface():
85
  css = """
86
- body {
87
- background: linear-gradient(to bottom, rgba(45, 11, 11, 0.9), rgba(0, 0, 0, 0.8)), url('/content/logo.jpg') no-repeat center center fixed;
88
- background-size: cover;
89
- min-height: 100vh;
90
- margin: 0;
91
- padding: 1rem;
92
- font-family: 'Poppins', sans-serif;
93
- color: #C0C0C0;
94
- overflow-x: hidden;
95
- }
96
- .header-text {
97
- text-align: center;
98
- padding: 100px 20px 20px;
99
- color: #ff4040;
100
- font-size: 3rem;
101
- font-weight: 900;
102
- text-shadow: 0 0 10px rgba(255, 64, 64, 0.5);
103
- z-index: 1500;
104
- animation: text-glow 2s infinite;
105
- }
106
- .header-subtitle {
107
- text-align: center;
108
- color: #C0C0C0;
109
- font-size: 1.2rem;
110
- font-weight: 300;
111
- margin-top: -10px;
112
- text-shadow: 0 0 5px rgba(255, 64, 64, 0.3);
113
- }
114
- .gr-tab {
115
- background: rgba(128, 0, 0, 0.5) !important;
116
- border-radius: 12px 12px 0 0 !important;
117
- margin: 0 5px !important;
118
- color: #C0C0C0 !important;
119
- border: 1px solid #ff4040 !important;
120
- z-index: 1500;
121
- transition: background 0.3s ease, color 0.3s ease;
122
- padding: 10px 20px !important;
123
- font-size: 1.1rem !important;
124
- }
125
- button {
126
- transition: all 0.3s cubic-bezier(0.4, 0, 0.2, 1) !important;
127
- background: #800000 !important;
128
- border: 1px solid #ff4040 !important;
129
- color: #C0C0C0 !important;
130
- border-radius: 8px !important;
131
- padding: 8px 16px !important;
132
- box-shadow: 0 2px 10px rgba(255, 64, 64, 0.3);
133
- }
134
- button:hover {
135
- transform: scale(1.05) !important;
136
- box-shadow: 0 10px 40px rgba(255, 64, 64, 0.7) !important;
137
- background: #ff4040 !important;
138
- }
139
- .compact-upload.horizontal {
140
- display: inline-flex !important;
141
- align-items: center !important;
142
- gap: 8px !important;
143
- max-width: 400px !important;
144
- height: 40px !important;
145
- padding: 0 12px !important;
146
- border: 1px solid #ff4040 !important;
147
- background: rgba(128, 0, 0, 0.5) !important;
148
- border-radius: 8px !important;
149
- }
150
- .compact-dropdown {
151
- --padding: 8px 12px !important;
152
- --radius: 10px !important;
153
- border: 1px solid #ff4040 !important;
154
- background: rgba(128, 0, 0, 0.5) !important;
155
- color: #C0C0C0 !important;
156
- }
157
- #custom-progress {
158
- margin-top: 10px;
159
- padding: 10px;
160
- background: rgba(128, 0, 0, 0.3);
161
- border-radius: 8px;
162
- border: 1px solid #ff4040;
163
- }
164
- #progress-bar {
165
- height: 20px;
166
- background: linear-gradient(to right, #6e8efb, #ff4040);
167
- border-radius: 5px;
168
- transition: width 0.5s ease-in-out;
169
- max-width: 100% !important;
170
- }
171
- .gr-accordion {
172
- background: rgba(128, 0, 0, 0.5) !important;
173
- border-radius: 10px !important;
174
- border: 1px solid #ff4040 !important;
175
- }
176
- .footer {
177
- text-align: center;
178
- padding: 20px;
179
- color: #ff4040;
180
- font-size: 14px;
181
- margin-top: 40px;
182
- background: rgba(128, 0, 0, 0.3);
183
- border-top: 1px solid #ff4040;
184
- }
185
- #log-accordion {
186
- max-height: 400px;
187
- overflow-y: auto;
188
- background: rgba(0, 0, 0, 0.7) !important;
189
- padding: 10px;
190
- border-radius: 8px;
191
- }
192
- @keyframes text-glow {
193
- 0% { text-shadow: 0 0 5px rgba(192, 192, 192, 0); }
194
- 50% { text-shadow: 0 0 15px rgba(192, 192, 192, 1); }
195
- 100% { text-shadow: 0 0 5px rgba(192, 192, 192, 0); }
196
- }
197
  """
198
 
199
- # Load user config
200
- user_config = load_config()
201
- initial_settings = user_config["settings"]
202
- initial_favorites = user_config["favorites"]
203
- initial_presets = user_config["presets"]
204
-
205
  with gr.Blocks(theme=gr.themes.Soft(), css=css) as demo:
206
- # Use hidden Textbox components to store state instead of gr.State
207
- current_lang = gr.Textbox(value=i18n.language, visible=False)
208
  favorites_state = gr.Textbox(value=json.dumps(initial_favorites), visible=False)
209
- presets_state = gr.Textbox(value=json.dumps(initial_presets), visible=False)
210
-
211
- header_html = gr.HTML(
212
- value=f"""
213
- <div class="header-text">{i18n("SESA Audio Separation")}</div>
214
- <div class="header-subtitle">{i18n("ultimate_audio_separation")}</div>
215
- """
216
- )
217
-
218
- with gr.Tabs():
219
- with gr.Tab(i18n("audio_separation_tab"), id="separation_tab"):
220
- with gr.Row(equal_height=True):
221
- with gr.Column(scale=1, min_width=380):
222
- with gr.Accordion(i18n("input_model"), open=True):
223
- with gr.Tabs():
224
- with gr.Tab(i18n("upload")):
225
- input_audio_file = gr.File(
226
- file_types=[".wav", ".mp3", ".m4a", ".mp4", ".mkv", ".flac"],
227
- elem_classes=["compact-upload", "horizontal"],
228
- label=i18n("upload_file")
229
- )
230
- with gr.Tab(i18n("path")):
231
- file_path_input = gr.Textbox(placeholder=i18n("path_placeholder"), label=i18n("file_path"))
232
-
233
- with gr.Row():
234
- model_category = gr.Dropdown(
235
- label=i18n("category"),
236
- choices=[i18n(cat) for cat in MODEL_CONFIGS.keys()],
237
- value=i18n(initial_settings["model_category"])
238
- )
239
- favorite_button = gr.Button(i18n("add_favorite"), variant="secondary", scale=0)
240
-
241
- model_dropdown = gr.Dropdown(
242
- label=i18n("model"),
243
- choices=update_model_dropdown(i18n(initial_settings["model_category"]), favorites=initial_favorites)["choices"],
244
- value=initial_settings["selected_model"]
245
- )
246
-
247
- with gr.Accordion(i18n("settings"), open=False):
248
- with gr.Row():
249
- export_format = gr.Dropdown(
250
- label=i18n("format"),
251
- choices=['wav FLOAT', 'flac PCM_16', 'flac PCM_24'],
252
- value=initial_settings["export_format"]
253
- )
254
- chunk_size = gr.Dropdown(
255
- label=i18n("chunk_size"),
256
- choices=[352800, 485100],
257
- value=initial_settings["chunk_size"],
258
- info=i18n("chunk_size_info")
259
- )
260
-
261
- overlap = gr.Slider(
262
- minimum=2,
263
- maximum=50,
264
- step=1,
265
- label=i18n("overlap"),
266
- value=initial_settings["overlap"],
267
- info=i18n("overlap_info")
268
- )
269
-
270
- use_tta = gr.Checkbox(
271
- label=i18n("tta_boost"),
272
- info=i18n("tta_info"),
273
- value=initial_settings["use_tta"]
274
- )
275
-
276
- use_demud_phaseremix_inst = gr.Checkbox(
277
- label=i18n("phase_fix"),
278
- info=i18n("phase_fix_info"),
279
- value=initial_settings["use_demud_phaseremix_inst"]
280
- )
281
-
282
- extract_instrumental = gr.Checkbox(
283
- label=i18n("instrumental"),
284
- info=i18n("instrumental_info"),
285
- value=initial_settings["extract_instrumental"]
286
- )
287
-
288
- use_apollo = gr.Checkbox(
289
- label=i18n("enhance_with_apollo"),
290
- value=initial_settings["use_apollo"],
291
- info=i18n("apollo_enhancement_info")
292
- )
293
-
294
- with gr.Group(visible=initial_settings["use_apollo"]) as apollo_settings_group:
295
- with gr.Row():
296
- apollo_chunk_size = gr.Slider(
297
- label=i18n("apollo_chunk_size"),
298
- minimum=3,
299
- maximum=25,
300
- step=1,
301
- value=initial_settings["apollo_chunk_size"],
302
- info=i18n("apollo_chunk_size_info")
303
- )
304
- apollo_overlap = gr.Slider(
305
- label=i18n("apollo_overlap"),
306
- minimum=2,
307
- maximum=10,
308
- step=1,
309
- value=initial_settings["apollo_overlap"],
310
- info=i18n("apollo_overlap_info")
311
- )
312
-
313
- apollo_method = gr.Dropdown(
314
- label=i18n("apollo_processing_method"),
315
- choices=[i18n("normal_method"), i18n("mid_side_method")],
316
- value=i18n(initial_settings["apollo_method"])
317
- )
318
-
319
- apollo_normal_model_row = gr.Row(visible=initial_settings["apollo_method"] != "mid_side_method")
320
- apollo_normal_model = gr.Dropdown(
321
- label=i18n("apollo_normal_model"),
322
- choices=["MP3 Enhancer", "Lew Vocal Enhancer", "Lew Vocal Enhancer v2 (beta)", "Apollo Universal Model"],
323
- value=initial_settings["apollo_normal_model"],
324
- container=apollo_normal_model_row
325
- )
326
-
327
- apollo_midside_model_row = gr.Row(visible=initial_settings["apollo_method"] == "mid_side_method")
328
- apollo_midside_model = gr.Dropdown(
329
- label=i18n("apollo_mid_side_model"),
330
- choices=["MP3 Enhancer", "Lew Vocal Enhancer", "Lew Vocal Enhancer v2 (beta)", "Apollo Universal Model"],
331
- value=initial_settings["apollo_midside_model"],
332
- container=apollo_midside_model_row
333
- )
334
-
335
- use_matchering = gr.Checkbox(
336
- label=i18n("apply_matchering"),
337
- value=initial_settings.get("use_matchering", False),
338
- info=i18n("matchering_info")
339
- )
340
-
341
- with gr.Group(visible=initial_settings.get("use_matchering", True)) as matchering_settings_group:
342
- matchering_passes = gr.Slider(
343
- label=i18n("matchering_passes"),
344
- minimum=1,
345
- maximum=5,
346
- step=1,
347
- value=initial_settings.get("matchering_passes", 1),
348
- info=i18n("matchering_passes_info")
349
- )
350
-
351
- with gr.Row():
352
- process_btn = gr.Button(i18n("process"), variant="primary")
353
- clear_old_output_btn = gr.Button(i18n("reset"), variant="secondary")
354
- clear_old_output_status = gr.Textbox(label=i18n("status"), interactive=False)
355
-
356
- with gr.Column(scale=2, min_width=800):
357
- with gr.Tabs():
358
- with gr.Tab(i18n("main_tab")):
359
- original_audio = gr.Audio(label=i18n("original"), interactive=True)
360
- with gr.Row():
361
- vocals_audio = gr.Audio(label=i18n("vocals"), show_download_button=True)
362
- instrumental_audio = gr.Audio(label=i18n("instrumental_output"), show_download_button=True)
363
- other_audio = gr.Audio(label=i18n("other"), show_download_button=True)
364
-
365
- with gr.Tab(i18n("details_tab")):
366
- with gr.Row():
367
- male_audio = gr.Audio(label=i18n("male"))
368
- female_audio = gr.Audio(label=i18n("female"))
369
- speech_audio = gr.Audio(label=i18n("speech"))
370
- with gr.Row():
371
- drum_audio = gr.Audio(label=i18n("drums"))
372
- bass_audio = gr.Audio(label=i18n("bass"))
373
- with gr.Row():
374
- effects_audio = gr.Audio(label=i18n("effects"))
375
-
376
- with gr.Tab(i18n("advanced_tab")):
377
- with gr.Row():
378
- phaseremix_audio = gr.Audio(label=i18n("phase_remix"))
379
- dry_audio = gr.Audio(label=i18n("dry"))
380
- with gr.Row():
381
- music_audio = gr.Audio(label=i18n("music"))
382
- karaoke_audio = gr.Audio(label=i18n("karaoke"))
383
- bleed_audio = gr.Audio(label=i18n("bleed"))
384
-
385
- separation_progress_html = gr.HTML(
386
- value=f"""
387
- <div id="custom-progress">
388
- <div id="progress-label">{i18n("waiting_for_processing")}</div>
389
- <div style="width: 100%; background-color: #444; border-radius: 5px;">
390
- <div id="progress-bar" style="width: 0%;"></div>
391
- </div>
392
- </div>
393
- """
394
- )
395
- separation_process_status = gr.Textbox(
396
- label=i18n("status"),
397
- interactive=False,
398
- placeholder=i18n("waiting_for_processing"),
399
- visible=False
400
- )
401
- processing_tip = gr.Markdown(i18n("processing_tip"))
402
-
403
- with gr.Tab(i18n("auto_ensemble_tab"), id="auto_ensemble_tab"):
404
- with gr.Row():
405
- with gr.Column():
406
- with gr.Group():
407
- auto_input_audio_file = gr.File(
408
- file_types=[".wav", ".mp3", ".m4a", ".mp4", ".mkv", ".flac"],
409
- label=i18n("upload_file")
410
- )
411
- auto_file_path_input = gr.Textbox(
412
- label=i18n("enter_file_path"),
413
- placeholder=i18n("file_path_placeholder")
414
- )
415
-
416
- with gr.Accordion(i18n("advanced_settings"), open=False):
417
- with gr.Row():
418
- auto_use_tta = gr.Checkbox(label=i18n("use_tta"), value=False)
419
- auto_extract_instrumental = gr.Checkbox(label=i18n("instrumental_only"))
420
-
421
- with gr.Row():
422
- auto_overlap = gr.Slider(
423
- label=i18n("auto_overlap"),
424
- minimum=2,
425
- maximum=50,
426
- value=2,
427
- step=1
428
- )
429
- auto_chunk_size = gr.Dropdown(
430
- label=i18n("auto_chunk_size"),
431
- choices=[352800, 485100],
432
- value=352800
433
- )
434
- export_format2 = gr.Dropdown(
435
- label=i18n("output_format"),
436
- choices=['wav FLOAT', 'flac PCM_16', 'flac PCM_24'],
437
- value='wav FLOAT'
438
- )
439
-
440
- auto_use_apollo = gr.Checkbox(
441
- label=i18n("enhance_with_apollo"),
442
- value=False,
443
- info=i18n("apollo_enhancement_info")
444
- )
445
-
446
- with gr.Group(visible=False) as auto_apollo_settings_group:
447
- with gr.Row():
448
- auto_apollo_chunk_size = gr.Slider(
449
- label=i18n("apollo_chunk_size"),
450
- minimum=3,
451
- maximum=25,
452
- step=1,
453
- value=19,
454
- info=i18n("apollo_chunk_size_info")
455
- )
456
- auto_apollo_overlap = gr.Slider(
457
- label=i18n("apollo_overlap"),
458
- minimum=2,
459
- maximum=10,
460
- step=1,
461
- value=2,
462
- info=i18n("apollo_overlap_info")
463
- )
464
-
465
- auto_apollo_method = gr.Dropdown(
466
- label=i18n("apollo_processing_method"),
467
- choices=[i18n("normal_method"), i18n("mid_side_method")],
468
- value=i18n("normal_method")
469
- )
470
-
471
- auto_apollo_normal_model_row = gr.Row(visible=True)
472
- auto_apollo_normal_model = gr.Dropdown(
473
- label=i18n("apollo_normal_model"),
474
- choices=["MP3 Enhancer", "Lew Vocal Enhancer", "Lew Vocal Enhancer v2 (beta)", "Apollo Universal Model"],
475
- value="Apollo Universal Model",
476
- container=auto_apollo_normal_model_row
477
- )
478
-
479
- auto_apollo_midside_model_row = gr.Row(visible=False)
480
- auto_apollo_midside_model = gr.Dropdown(
481
- label=i18n("apollo_mid_side_model"),
482
- choices=["MP3 Enhancer", "Lew Vocal Enhancer", "Lew Vocal Enhancer v2 (beta)", "Apollo Universal Model"],
483
- value="Apollo Universal Model",
484
- container=auto_apollo_midside_model_row
485
- )
486
-
487
- auto_use_matchering = gr.Checkbox(
488
- label=i18n("apply_matchering"),
489
- value=False,
490
- info=i18n("matchering_info")
491
- )
492
-
493
- with gr.Group(visible=False) as auto_matchering_settings_group:
494
- auto_matchering_passes = gr.Slider(
495
- label=i18n("matchering_passes"),
496
- minimum=1,
497
- maximum=5,
498
- step=1,
499
- value=1,
500
- info=i18n("matchering_passes_info")
501
- )
502
-
503
- with gr.Group():
504
- model_selection_header = gr.Markdown(f"### {i18n('model_selection')}")
505
- with gr.Row():
506
- auto_category_dropdown = gr.Dropdown(
507
- label=i18n("model_category"),
508
- choices=[i18n(cat) for cat in MODEL_CONFIGS.keys()],
509
- value=i18n("Vocal Models")
510
- )
511
- selected_models = gr.Dropdown(
512
- label=i18n("selected_models"),
513
- choices=update_model_dropdown(i18n(initial_settings["auto_category"]), favorites=initial_favorites)["choices"],
514
- value=initial_settings["selected_models"],
515
- multiselect=True
516
- )
517
-
518
- preset_dropdown = gr.Dropdown(
519
- label=i18n("select_preset"),
520
- choices=list(initial_presets.keys()),
521
- value=None,
522
- allow_custom_value=False
523
- )
524
- with gr.Row():
525
- preset_name_input = gr.Textbox(
526
- label=i18n("preset_name"),
527
- placeholder=i18n("enter_preset_name")
528
- )
529
- save_preset_btn = gr.Button(i18n("save_preset"), variant="secondary", scale=0)
530
- delete_preset_btn = gr.Button(i18n("delete_preset"), variant="secondary", scale=0)
531
- refresh_presets_btn = gr.Button(i18n("refresh_presets"), variant="secondary", scale=0)
532
-
533
- with gr.Group():
534
- ensemble_settings_header = gr.Markdown(f"### {i18n('ensemble_settings')}")
535
- auto_ensemble_type = gr.Dropdown(
536
- label=i18n("method"),
537
- choices=['avg_wave', 'median_wave', 'min_wave', 'max_wave',
538
- 'avg_fft', 'median_fft', 'min_fft', 'max_fft'],
539
- value=initial_settings["auto_ensemble_type"]
540
- )
541
-
542
- ensemble_recommendation = gr.Markdown(i18n("recommendation"))
543
-
544
- auto_process_btn = gr.Button(i18n("start_processing"), variant="primary")
545
-
546
- with gr.Column():
547
- with gr.Tabs():
548
- with gr.Tab(i18n("original_audio_tab")):
549
- original_audio2 = gr.Audio(
550
- label=i18n("original_audio"),
551
- interactive=True,
552
- elem_id="original_audio_player"
553
- )
554
- with gr.Tab(i18n("ensemble_result_tab")):
555
- auto_output_audio = gr.Audio(
556
- label=i18n("output_preview"),
557
- show_download_button=True,
558
- interactive=True
559
- )
560
- refresh_output_btn = gr.Button(i18n("refresh_output"), variant="secondary")
561
-
562
- ensemble_progress_html = gr.HTML(
563
- value=f"""
564
- <div id="custom-progress">
565
- <div id="progress-label">{i18n("waiting_for_processing")}</div>
566
- <div style="width: 100%; background-color: #444; border-radius: 5px;">
567
- <div id="progress-bar" style="width: 0%;"></div>
568
- </div>
569
- </div>
570
- """
571
- )
572
- ensemble_process_status = gr.Textbox(
573
- label=i18n("status"),
574
- interactive=False,
575
- placeholder=i18n("waiting_for_processing"),
576
- visible=False
577
- )
578
-
579
- with gr.Tab(i18n("download_sources_tab"), id="download_tab"):
580
- with gr.Row():
581
- with gr.Column():
582
- gr.Markdown(f"### {i18n('direct_links')}")
583
- direct_url_input = gr.Textbox(label=i18n("audio_file_url"))
584
- direct_download_btn = gr.Button(i18n("download_from_url"), variant="secondary")
585
- direct_download_status = gr.Textbox(label=i18n("download_status"))
586
- direct_download_output = gr.File(label=i18n("downloaded_file"), interactive=False)
587
-
588
- with gr.Column():
589
- gr.Markdown(f"### {i18n('cookie_management')}")
590
- cookie_file = gr.File(
591
- label=i18n("upload_cookies_txt"),
592
- file_types=[".txt"],
593
- interactive=True,
594
- elem_id="cookie_upload"
595
- )
596
- cookie_info = gr.Markdown(i18n("cookie_info"))
597
-
598
- with gr.Tab(i18n("manual_ensemble_tab"), id="manual_ensemble_tab"):
599
- with gr.Row(equal_height=True):
600
- with gr.Column(scale=1, min_width=400):
601
- with gr.Accordion(i18n("input_sources"), open=True):
602
- with gr.Row():
603
- refresh_btn = gr.Button(i18n("refresh"), variant="secondary", size="sm")
604
- ensemble_type = gr.Dropdown(
605
- label=i18n("ensemble_algorithm"),
606
- choices=['avg_wave', 'median_wave', 'min_wave', 'max_wave',
607
- 'avg_fft', 'median_fft', 'min_fft', 'max_fft'],
608
- value='avg_wave'
609
- )
610
 
611
- file_dropdown_header = gr.Markdown(f"### {i18n('select_audio_files')}")
612
- file_path = os.path.join(Path.home(), 'Music-Source-Separation', 'output')
613
- initial_files = glob.glob(f"{file_path}/*.wav") + glob.glob(os.path.join(BASE_DIR, 'Music-Source-Separation-Training', 'old_output', '*.wav'))
614
- file_dropdown = gr.Dropdown(
615
- choices=initial_files,
616
- label=i18n("available_files"),
617
- multiselect=True,
618
- interactive=True,
619
- elem_id="file-dropdown"
620
- )
621
- weights_input = gr.Textbox(
622
- label=i18n("custom_weights"),
623
- placeholder=i18n("custom_weights_placeholder"),
624
- info=i18n("custom_weights_info")
625
- )
626
-
627
- with gr.Column(scale=2, min_width=800):
628
- with gr.Tabs():
629
- with gr.Tab(i18n("result_preview_tab")):
630
- ensemble_output_audio = gr.Audio(
631
- label=i18n("ensembled_output"),
632
- interactive=True,
633
- show_download_button=True,
634
- elem_id="output-audio"
635
- )
636
- with gr.Tab(i18n("processing_log_tab")):
637
- with gr.Accordion(i18n("processing_details"), open=True, elem_id="log-accordion"):
638
- ensemble_status = gr.Textbox(
639
- label="",
640
- interactive=False,
641
- placeholder=i18n("processing_log_placeholder"),
642
- lines=10,
643
- max_lines=20,
644
- elem_id="log-box"
645
- )
646
- ensemble_process_btn = gr.Button(
647
- i18n("process_ensemble"),
648
- variant="primary",
649
- size="sm",
650
- elem_id="process-btn"
651
- )
652
 
653
  # Event handlers
654
- def update_favorite_button(model, favorites_json):
655
- favorites = json.loads(favorites_json)
656
- cleaned_model = clean_model(model) if model else None
657
- is_favorited = cleaned_model in favorites if cleaned_model else False
658
- return gr.update(value=i18n("remove_favorite") if is_favorited else i18n("add_favorite"))
659
-
660
- def toggle_favorite(model, favorites_json):
661
- favorites = json.loads(favorites_json)
662
- if not model:
663
- return json.dumps(favorites), gr.update(), gr.update()
664
- cleaned_model = clean_model(model)
665
- is_favorited = cleaned_model in favorites
666
- new_favorites = update_favorites(favorites, cleaned_model, add=not is_favorited)
667
- save_config(new_favorites, load_config()["settings"], load_config()["presets"])
668
- category = model_category.value
669
- return (
670
- json.dumps(new_favorites),
671
- gr.update(choices=update_model_dropdown(category, favorites=new_favorites)["choices"]),
672
- gr.update(value=i18n("add_favorite") if is_favorited else i18n("remove_favorite"))
673
- )
674
-
675
- model_dropdown.change(
676
- fn=update_favorite_button,
677
- inputs=[model_dropdown, favorites_state],
678
- outputs=favorite_button
679
- )
680
-
681
- favorite_button.click(
682
- fn=toggle_favorite,
683
- inputs=[model_dropdown, favorites_state],
684
- outputs=[favorites_state, model_dropdown, favorite_button]
685
- )
686
-
687
- use_apollo.change(
688
- fn=lambda x: gr.update(visible=x),
689
- inputs=use_apollo,
690
- outputs=apollo_settings_group
691
- )
692
-
693
- use_matchering.change(
694
- fn=lambda x: gr.update(visible=x),
695
- inputs=use_matchering,
696
- outputs=matchering_settings_group
697
- )
698
-
699
- apollo_method.change(
700
- fn=lambda x: [
701
- gr.update(visible=x != i18n("mid_side_method")),
702
- gr.update(visible=x == i18n("mid_side_method")),
703
- "Apollo Universal Model" if x == i18n("mid_side_method") else None
704
- ],
705
- inputs=apollo_method,
706
- outputs=[apollo_normal_model_row, apollo_midside_model_row, apollo_normal_model]
707
- )
708
-
709
- def load_preset(preset_name, presets_json, category, favorites_json):
710
- presets = json.loads(presets_json)
711
- favorites = json.loads(favorites_json)
712
- if preset_name and preset_name in presets:
713
- preset = presets[preset_name]
714
- favorite_models = [f"{model} ⭐" if model in favorites else model for model in preset["models"]]
715
- preset_category = preset.get("auto_category_dropdown", category)
716
- model_choices = update_model_dropdown(preset_category, favorites=favorites)["choices"]
717
- logging.debug(f"Preset '{preset_name}' loaded with models: {favorite_models}, category: {preset_category}")
718
- return (
719
- gr.update(value=preset_category),
720
- gr.update(choices=model_choices, value=favorite_models),
721
- gr.update(value=preset["ensemble_method"])
722
- )
723
- logging.debug(f"Preset '{preset_name}' not found.")
724
- return gr.update(), gr.update(), gr.update()
725
-
726
- def sync_presets():
727
- config = load_config()
728
- return json.dumps(config["presets"]), gr.update(choices=list(config["presets"].keys()), value=None)
729
-
730
- preset_dropdown.change(
731
- fn=load_preset,
732
- inputs=[preset_dropdown, presets_state, auto_category_dropdown, favorites_state],
733
- outputs=[auto_category_dropdown, selected_models, auto_ensemble_type]
734
- )
735
-
736
- def handle_save_preset(preset_name, models, ensemble_method, presets_json, favorites_json, auto_category_dropdown):
737
- presets = json.loads(presets_json)
738
- favorites = json.loads(favorites_json)
739
- if not preset_name:
740
- return gr.update(), json.dumps(presets), i18n("no_preset_name_provided")
741
- if not models and not favorites:
742
- return gr.update(), json.dumps(presets), i18n("no_models_selected_for_preset")
743
- new_presets = save_preset(
744
- presets,
745
- preset_name,
746
- models,
747
- ensemble_method,
748
- auto_category_dropdown=auto_category_dropdown
749
- )
750
- save_config(favorites, load_config()["settings"], new_presets)
751
- logging.debug(f"Preset dropdown updated with choices: {list(new_presets.keys())}")
752
- return gr.update(choices=list(new_presets.keys()), value=None), json.dumps(new_presets), i18n("preset_saved").format(preset_name)
753
-
754
- save_preset_btn.click(
755
- fn=handle_save_preset,
756
- inputs=[preset_name_input, selected_models, auto_ensemble_type, presets_state, favorites_state, auto_category_dropdown],
757
- outputs=[preset_dropdown, presets_state]
758
- )
759
-
760
- def handle_delete_preset(preset_name, presets_json):
761
- presets = json.loads(presets_json)
762
- if not preset_name or preset_name not in presets:
763
- return gr.update(), json.dumps(presets)
764
- new_presets = delete_preset(presets, preset_name)
765
- save_config(load_config()["favorites"], load_config()["settings"], new_presets)
766
- return gr.update(choices=list(new_presets.keys()), value=None), json.dumps(new_presets)
767
-
768
- delete_preset_btn.click(
769
- fn=handle_delete_preset,
770
- inputs=[preset_dropdown, presets_state],
771
- outputs=[preset_dropdown, presets_state]
772
- )
773
-
774
- refresh_presets_btn.click(
775
- fn=sync_presets,
776
- inputs=[],
777
- outputs=[presets_state, preset_dropdown]
778
- )
779
-
780
- auto_use_apollo.change(
781
- fn=lambda x: gr.update(visible=x),
782
- inputs=auto_use_apollo,
783
- outputs=auto_apollo_settings_group
784
- )
785
-
786
- auto_use_matchering.change(
787
- fn=lambda x: gr.update(visible=x),
788
- inputs=auto_use_matchering,
789
- outputs=auto_matchering_settings_group
790
- )
791
-
792
- auto_apollo_method.change(
793
- fn=lambda x: [
794
- gr.update(visible=x != i18n("mid_side_method")),
795
- gr.update(visible=x == i18n("mid_side_method")),
796
- "Apollo Universal Model" if x == i18n("mid_side_method") else None
797
- ],
798
- inputs=auto_apollo_method,
799
- outputs=[auto_apollo_normal_model_row, auto_apollo_midside_model_row, auto_apollo_normal_model]
800
- )
801
-
802
- def save_settings_on_process(*args):
803
- apollo_method_value = args[11]
804
- backend_apollo_method = "mid_side_method" if apollo_method_value == i18n("mid_side_method") else "normal_method"
805
- cleaned_model = clean_model(args[1]) if args[1] else None
806
- settings = {
807
- "chunk_size": args[2],
808
- "overlap": args[3],
809
- "export_format": args[4],
810
- "use_tta": args[5],
811
- "use_demud_phaseremix_inst": args[6],
812
- "extract_instrumental": args[7],
813
- "use_apollo": args[8],
814
- "apollo_chunk_size": args[9],
815
- "apollo_overlap": args[10],
816
- "apollo_method": backend_apollo_method,
817
- "apollo_normal_model": args[12],
818
- "apollo_midside_model": args[13],
819
- "use_matchering": args[14],
820
- "matchering_passes": args[15],
821
- "model_category": args[16],
822
- "selected_model": cleaned_model,
823
- "auto_ensemble_type": args[7]
824
- }
825
- save_config(load_config()["favorites"], settings, load_config()["presets"])
826
- modified_args = list(args)
827
- modified_args[1] = cleaned_model
828
- modified_args[17] = cleaned_model
829
- return process_audio(*modified_args)
830
-
831
- def save_auto_ensemble_settings(*args):
832
- settings = load_config()["settings"]
833
- settings["auto_ensemble_type"] = args[7]
834
- settings["use_matchering"] = args[14]
835
- settings["matchering_passes"] = args[15]
836
- save_config(load_config()["favorites"], settings, load_config()["presets"])
837
- output_audio, status, progress_html = None, i18n("waiting_for_processing"), ensemble_progress_html.value
838
- for update in auto_ensemble_process(*args):
839
- if isinstance(update, tuple) and len(update) == 3:
840
- output_audio, status, progress_html = update
841
- return output_audio, status, progress_html
842
-
843
- def update_category_dropdowns(cat):
844
- logging.debug(f"Input category: {cat}")
845
- eng_cat = next((k for k in MODEL_CONFIGS.keys() if i18n(k) == cat), list(MODEL_CONFIGS.keys())[0])
846
- logging.debug(f"Using English category: {eng_cat}")
847
- choices = update_model_dropdown(eng_cat, favorites=load_config()["favorites"])["choices"]
848
- logging.debug(f"Model choices: {choices}")
849
- return gr.update(choices=choices), gr.update(choices=choices)
850
 
851
  model_category.change(
852
- fn=update_category_dropdowns,
853
  inputs=model_category,
854
- outputs=[model_dropdown, selected_models]
855
  )
856
 
857
- clear_old_output_btn.click(fn=clear_old_output, outputs=clear_old_output_status)
858
-
859
  input_audio_file.upload(
860
- fn=lambda x, y: handle_file_upload(x, y, is_auto_ensemble=False),
861
- inputs=[input_audio_file, file_path_input],
862
- outputs=[input_audio_file, original_audio]
863
- )
864
- file_path_input.change(
865
- fn=lambda x, y: handle_file_upload(x, y, is_auto_ensemble=False),
866
- inputs=[input_audio_file, file_path_input],
867
  outputs=[input_audio_file, original_audio]
868
  )
869
 
870
- auto_input_audio_file.upload(
871
- fn=lambda x, y: handle_file_upload(x, y, is_auto_ensemble=True),
872
- inputs=[auto_input_audio_file, auto_file_path_input],
873
- outputs=[auto_input_audio_file, original_audio2]
874
- )
875
- auto_file_path_input.change(
876
- fn=lambda x, y: handle_file_upload(x, y, is_auto_ensemble=True),
877
- inputs=[auto_input_audio_file, auto_file_path_input],
878
- outputs=[auto_input_audio_file, original_audio2]
879
- )
880
-
881
- auto_category_dropdown.change(
882
- fn=lambda cat: gr.update(choices=update_model_dropdown(next((k for k in MODEL_CONFIGS.keys() if i18n(k) == cat), list(MODEL_CONFIGS.keys())[0]), favorites=load_config()["favorites"])["choices"]),
883
- inputs=auto_category_dropdown,
884
- outputs=selected_models
885
- )
886
-
887
- def debug_inputs(*args):
888
- input_names = [
889
- "input_audio_file", "model_dropdown", "chunk_size", "overlap", "export_format",
890
- "use_tta", "use_demud_phaseremix_inst", "extract_instrumental",
891
- "use_apollo", "apollo_chunk_size", "apollo_overlap",
892
- "apollo_method", "apollo_normal_model", "apollo_midside_model",
893
- "use_matchering", "matchering_passes", "model_category", "selected_model"
894
- ]
895
- cleaned_args = list(args)
896
- cleaned_args[1] = clean_model(cleaned_args[1]) if cleaned_args[1] else None
897
- cleaned_args[17] = clean_model(cleaned_args[17]) if cleaned_args[17] else None
898
- for name, value in zip(input_names, cleaned_args):
899
- logging.debug(f"UI Input - {name}: {value}")
900
- return args
901
-
902
  process_btn.click(
903
- fn=lambda *args: save_settings_on_process(*debug_inputs(*args)),
904
- inputs=[
905
- input_audio_file, model_dropdown, chunk_size, overlap, export_format,
906
- use_tta, use_demud_phaseremix_inst, extract_instrumental,
907
- use_apollo, apollo_chunk_size, apollo_overlap,
908
- apollo_method, apollo_normal_model, apollo_midside_model,
909
- use_matchering, matchering_passes, model_category, model_dropdown
910
- ],
911
- outputs=[
912
- vocals_audio, instrumental_audio, phaseremix_audio, drum_audio, karaoke_audio,
913
- other_audio, bass_audio, effects_audio, speech_audio, bleed_audio, music_audio,
914
- dry_audio, male_audio, female_audio,
915
- separation_process_status, separation_progress_html
916
- ]
917
  )
918
 
919
- auto_process_btn.click(
920
- fn=save_auto_ensemble_settings,
921
- inputs=[
922
- auto_input_audio_file,
923
- selected_models,
924
- auto_chunk_size,
925
- auto_overlap,
926
- export_format2,
927
- auto_use_tta,
928
- auto_extract_instrumental,
929
- auto_ensemble_type,
930
- gr.State(None),
931
- auto_use_apollo,
932
- auto_apollo_normal_model,
933
- auto_apollo_chunk_size,
934
- auto_apollo_overlap,
935
- auto_apollo_method,
936
- auto_use_matchering,
937
- auto_matchering_passes,
938
- auto_apollo_midside_model
939
- ],
940
- outputs=[auto_output_audio, ensemble_process_status, ensemble_progress_html]
941
  )
942
 
943
- direct_download_btn.click(
944
- fn=download_callback,
945
- inputs=[direct_url_input, gr.State('direct'), cookie_file],
946
- outputs=[direct_download_output, direct_download_status, input_audio_file, auto_input_audio_file, original_audio, original_audio2]
947
- )
948
-
949
- refresh_output_btn.click(
950
- fn=refresh_auto_output,
951
- inputs=[],
952
- outputs=[auto_output_audio, ensemble_process_status]
953
- )
954
-
955
- refresh_btn.click(fn=update_file_list, outputs=file_dropdown)
956
- ensemble_process_btn.click(fn=ensemble_audio_fn, inputs=[file_dropdown, ensemble_type, weights_input], outputs=[ensemble_output_audio, ensemble_status])
957
-
958
  return demo
 
1
  import gradio as gr
2
  import os
 
 
 
 
3
  import json
 
 
 
 
 
 
 
 
 
4
  import logging
5
+ from helpers import update_model_dropdown, handle_file_upload, clear_old_output
6
+ from model import MODEL_CONFIGS
7
+ from processing import process_audio
8
+ from assets.i18n.i18n import I18nAuto
9
+ from config_manager import load_config
10
 
11
  logging.basicConfig(filename='sesa_gui.log', level=logging.DEBUG)
12
 
13
+ # BASE_DIR and config setup
14
  BASE_DIR = os.path.dirname(os.path.abspath(__file__))
15
  CONFIG_DIR = os.path.join(BASE_DIR, "assets")
16
  CONFIG_FILE = os.path.join(CONFIG_DIR, "config.json")
 
17
 
18
+ # Load user config
19
  user_config = load_config()
20
  initial_settings = user_config["settings"]
21
  initial_favorites = user_config["favorites"]
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
22
 
23
  # Initialize I18nAuto
24
  i18n = I18nAuto()
25
 
 
 
 
 
26
  def create_interface():
27
  css = """
28
+ body { background: #222; color: #fff; font-family: Arial, sans-serif; }
29
+ .gr-tab { background: #444 !important; color: #fff !important; border: 1px solid #888 !important; }
30
+ button { background: #555 !important; color: #fff !important; border: 1px solid #888 !important; }
31
+ button:hover { background: #777 !important; }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
32
  """
33
 
 
 
 
 
 
 
34
  with gr.Blocks(theme=gr.themes.Soft(), css=css) as demo:
35
+ # Store state in hidden Textbox
 
36
  favorites_state = gr.Textbox(value=json.dumps(initial_favorites), visible=False)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
37
 
38
+ gr.HTML(f"<h1>{i18n('SESA Audio Separation')}</h1>")
39
+
40
+ with gr.Tab(i18n("audio_separation_tab")):
41
+ with gr.Row():
42
+ with gr.Column():
43
+ input_audio_file = gr.File(
44
+ file_types=[".wav", ".mp3", ".flac"],
45
+ label=i18n("upload_file")
46
+ )
47
+ model_category = gr.Dropdown(
48
+ label=i18n("category"),
49
+ choices=[i18n(cat) for cat in MODEL_CONFIGS.keys()],
50
+ value=i18n(initial_settings["model_category"])
51
+ )
52
+ model_dropdown = gr.Dropdown(
53
+ label=i18n("model"),
54
+ choices=update_model_dropdown(i18n(initial_settings["model_category"]), favorites=initial_favorites)["choices"],
55
+ value=initial_settings["selected_model"]
56
+ )
57
+ export_format = gr.Dropdown(
58
+ label=i18n("format"),
59
+ choices=['wav FLOAT', 'flac PCM_16'],
60
+ value=initial_settings["export_format"]
61
+ )
62
+ process_btn = gr.Button(i18n("process"))
63
+ clear_old_output_btn = gr.Button(i18n("reset"))
64
+
65
+ with gr.Column():
66
+ original_audio = gr.Audio(label=i18n("original"))
67
+ vocals_audio = gr.Audio(label=i18n("vocals"))
68
+ instrumental_audio = gr.Audio(label=i18n("instrumental_output"))
69
+ status = gr.Textbox(label=i18n("status"), interactive=False)
 
 
 
 
 
 
 
 
 
70
 
71
  # Event handlers
72
+ def update_models(category):
73
+ logging.debug(f"Updating models for category: {category}")
74
+ choices = update_model_dropdown(category, favorites=json.loads(favorites_state.value))["choices"]
75
+ return gr.update(choices=choices)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
76
 
77
  model_category.change(
78
+ fn=update_models,
79
  inputs=model_category,
80
+ outputs=model_dropdown
81
  )
82
 
 
 
83
  input_audio_file.upload(
84
+ fn=lambda x: handle_file_upload(x, None, is_auto_ensemble=False),
85
+ inputs=[input_audio_file],
 
 
 
 
 
86
  outputs=[input_audio_file, original_audio]
87
  )
88
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
89
  process_btn.click(
90
+ fn=process_audio,
91
+ inputs=[input_audio_file, model_dropdown, export_format],
92
+ outputs=[vocals_audio, instrumental_audio, status]
 
 
 
 
 
 
 
 
 
 
 
93
  )
94
 
95
+ clear_old_output_btn.click(
96
+ fn=clear_old_output,
97
+ outputs=status
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
98
  )
99
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
100
  return demo