ASesYusuf1/SESA_Audio_Separation
14
1import gradio as gr2import os3import glob4import subprocess5from pathlib import Path6from datetime import datetime7import json8import sys9import time10import random11from helpers import update_model_dropdown, handle_file_upload, clear_old_output, save_uploaded_file, update_file_list, clean_model, get_model_categories12from download import download_callback13from model import get_model_config, MODEL_CONFIGS, get_all_model_configs_with_custom, add_custom_model, delete_custom_model, get_custom_models_list, SUPPORTED_MODEL_TYPES, load_custom_models, get_model_chunk_size14from processing import process_audio, auto_ensemble_process, ensemble_audio_fn, refresh_auto_output15from assets.i18n.i18n import I18nAuto16from config_manager import load_config, save_config, update_favorites, save_preset, delete_preset17from phase_fixer import SOURCE_MODELS, TARGET_MODELS18import logging19logging.basicConfig(filename='sesa_gui.log', level=logging.WARNING)20 21# BASE_DIR tanımı22BASE_DIR = os.path.dirname(os.path.abspath(__file__))23CONFIG_DIR = os.path.join(BASE_DIR, "assets")24CONFIG_FILE = os.path.join(CONFIG_DIR, "config.json")25URL_FILE = os.path.join(CONFIG_DIR, "last_url.txt")26 27# Load user config at startup28user_config = load_config()29initial_settings = user_config["settings"]30initial_favorites = user_config["favorites"]31initial_presets = user_config["presets"]32 33# Ensure auto_category is valid34if "auto_category" not in initial_settings or initial_settings["auto_category"] not in MODEL_CONFIGS:35 initial_settings["auto_category"] = "Vocal Models"36 37# Config dosyası yoksa oluştur38if 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) # Random port instead of fixed45 }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)50else: # If the file exists, load and update if necessary51 try:52 with open(CONFIG_FILE, "r", encoding="utf-8") as f:53 config = json.load(f)54 # Ensure 'lang' key exists55 if "lang" not in config:56 config["lang"] = {"override": False, "selected_lang": "auto"}57 # Add 'sharing' key if it doesn't exist58 if "sharing" not in config:59 config["sharing"] = {60 "method": "gradio",61 "ngrok_token": "",62 "port": random.randint(1000, 9000) # Random port instead of fixed63 }64 # Save the updated configuration65 with open(CONFIG_FILE, "w", encoding="utf-8") as f:66 json.dump(config, f, indent=2)67 except json.JSONDecodeError: # Handle corrupted JSON68 print("Warning: config.json is corrupted. Creating a new one.")69 default_config = {70 "lang": {"override": False, "selected_lang": "auto"},71 "sharing": {72 "method": "gradio",73 "ngrok_token": "",74 "port": random.randint(1000, 9000) # Random port instead of fixed75 }76 }77 with open(CONFIG_FILE, "w", encoding="utf-8") as f:78 json.dump(default_config, f, indent=2)79 80# I18nAuto örneği (arayüz başlamadan önce dil yüklenir)81i18n = I18nAuto()82 83# Çıktı formatları84OUTPUT_FORMATS = ['wav', 'flac', 'mp3', 'ogg', 'opus', 'm4a', 'aiff', 'ac3']85 86# Arayüz oluşturma fonksiyonu87def create_interface():88 css = """89 body {90 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;91 background-size: cover;92 min-height: 100vh;93 margin: 0;94 padding: 1rem;95 font-family: 'Poppins', sans-serif;96 color: #C0C0C0;97 overflow-x: hidden;98 }99 .header-text {100 text-align: center;101 padding: 100px 20px 20px;102 color: #ff4040;103 font-size: 3rem;104 font-weight: 900;105 text-shadow: 0 0 10px rgba(255, 64, 64, 0.5);106 z-index: 1500;107 animation: text-glow 2s infinite;108 }109 .header-subtitle {110 text-align: center;111 color: #C0C0C0;112 font-size: 1.2rem;113 font-weight: 300;114 margin-top: -10px;115 text-shadow: 0 0 5px rgba(255, 64, 64, 0.3);116 }117 .gr-tab {118 background: rgba(128, 0, 0, 0.5) !important;119 border-radius: 12px 12px 0 0 !important;120 margin: 0 5px !important;121 color: #C0C0C0 !important;122 border: 1px solid #ff4040 !important;123 z-index: 1500;124 transition: background 0.3s ease, color 0.3s ease;125 padding: 10px 20px !important;126 font-size: 1.1rem !important;127 }128 button {129 transition: all 0.3s cubic-bezier(0.4, 0, 0.2, 1) !important;130 background: #800000 !important;131 border: 1px solid #ff4040 !important;132 color: #C0C0C0 !important;133 border-radius: 8px !important;134 padding: 8px 16px !important;135 box-shadow: 0 2px 10px rgba(255, 64, 64, 0.3);136 }137 button:hover {138 transform: scale(1.05) !important;139 box-shadow: 0 10px 40px rgba(255, 64, 64, 0.7) !important;140 background: #ff4040 !important;141 }142 .compact-upload.horizontal {143 display: inline-flex !important;144 align-items: center !important;145 gap: 8px !important;146 max-width: 400px !important;147 height: 40px !important;148 padding: 0 12px !important;149 border: 1px solid #ff4040 !important;150 background: rgba(128, 0, 0, 0.5) !important;151 border-radius: 8px !important;152 }153 .compact-dropdown {154 --padding: 8px 12px !important;155 --radius: 10px !important;156 border: 1px solid #ff4040 !important;157 background: rgba(128, 0, 0, 0.5) !important;158 color: #C0C0C0 !important;159 }160 #custom-progress {161 margin-top: 10px;162 padding: 10px;163 background: rgba(128, 0, 0, 0.3);164 border-radius: 8px;165 border: 1px solid #ff4040;166 }167 #progress-bar {168 height: 20px;169 background: linear-gradient(90deg, #6e8efb, #a855f7, #ff4040);170 background-size: 200% 100%;171 border-radius: 5px;172 transition: width 0.4s cubic-bezier(0.4, 0, 0.2, 1);173 max-width: 100% !important;174 }175 @keyframes progress-shimmer {176 0% { background-position: 200% 0; }177 100% { background-position: -200% 0; }178 }179 #progress-bar[data-active="true"] {180 animation: progress-shimmer 2s linear infinite;181 }182 .gr-accordion {183 background: rgba(128, 0, 0, 0.5) !important;184 border-radius: 10px !important;185 border: 1px solid #ff4040 !important;186 }187 .footer {188 text-align: center;189 padding: 20px;190 color: #ff4040;191 font-size: 14px;192 margin-top: 40px;193 background: rgba(128, 0, 0, 0.3);194 border-top: 1px solid #ff4040;195 }196 #log-accordion {197 max-height: 400px;198 overflow-y: auto;199 background: rgba(0, 0, 0, 0.7) !important;200 padding: 10px;201 border-radius: 8px;202 }203 @keyframes text-glow {204 0% { text-shadow: 0 0 5px rgba(192, 192, 192, 0); }205 50% { text-shadow: 0 0 15px rgba(192, 192, 192, 1); }206 100% { text-shadow: 0 0 5px rgba(192, 192, 192, 0); }207 }208 """209 210 # Load user config at startup211 user_config = load_config()212 initial_settings = user_config["settings"]213 initial_favorites = user_config["favorites"]214 initial_presets = user_config["presets"]215 216 with gr.Blocks(theme=gr.themes.Soft(), css=css) as demo:217 current_lang = gr.State(value=i18n.language)218 favorites_state = gr.State(value=initial_favorites)219 presets_state = gr.State(value=initial_presets)220 221 header_html = gr.HTML(222 value=f"""223 <div class="header-text">{i18n("SESA Audio Separation")}</div>224 <div class="header-subtitle">{i18n("ultimate_audio_separation")}</div>225 """226 )227 228 with gr.Tabs():229 with gr.Tab(i18n("audio_separation_tab"), id="separation_tab"):230 with gr.Row(equal_height=True):231 with gr.Column(scale=1, min_width=380):232 with gr.Accordion(i18n("input_model"), open=True) as input_model_accordion:233 with gr.Tabs():234 with gr.Tab(i18n("upload")) as upload_tab:235 input_audio_file = gr.File(236 file_types=[".wav", ".mp3", ".m4a", ".mp4", ".mkv", ".flac"],237 elem_classes=["compact-upload", "horizontal", "x-narrow"],238 label=""239 )240 with gr.Tab(i18n("path")) as path_tab:241 file_path_input = gr.Textbox(placeholder=i18n("path_placeholder"))242 243 with gr.Row():244 model_category = gr.Dropdown(245 label=i18n("category"),246 choices=[i18n(cat) for cat in get_all_model_configs_with_custom().keys()],247 value=i18n(initial_settings["model_category"])248 )249 favorite_button = gr.Button(i18n("add_favorite"), variant="secondary", scale=0)250 251 model_dropdown = gr.Dropdown(252 label=i18n("model"),253 choices=update_model_dropdown(i18n(initial_settings["model_category"]), favorites=initial_favorites)["choices"],254 value=initial_settings["selected_model"]255 )256 257 with gr.Accordion(i18n("settings"), open=False) as settings_accordion:258 with gr.Row():259 with gr.Column(scale=1):260 export_format = gr.Dropdown(261 label=i18n("format"),262 choices=['wav FLOAT', 'flac PCM_16', 'flac PCM_24'],263 value=initial_settings["export_format"]264 )265 with gr.Column(scale=1):266 _init_cs_mode = initial_settings.get("chunk_size_mode", "base")267 chunk_size_mode = gr.Radio(268 label=i18n("chunk_size_mode"),269 choices=["base", "custom", "yaml"],270 value=_init_cs_mode,271 info=i18n("chunk_size_mode_info")272 )273 chunk_size = gr.Dropdown(274 label=i18n("chunk_size"),275 choices=[352800, 485100],276 value=initial_settings["chunk_size"],277 info=i18n("chunk_size_info"),278 visible=(_init_cs_mode == "base")279 )280 chunk_size_custom = gr.Number(281 label=i18n("chunk_size_custom_label"),282 value=initial_settings.get("chunk_size_custom", 352800),283 precision=0,284 info=i18n("chunk_size_custom_info"),285 visible=(_init_cs_mode == "custom")286 )287 chunk_size_yaml_display = gr.Textbox(288 label=i18n("chunk_size_yaml_label"),289 value=i18n("chunk_size_yaml_not_downloaded"),290 interactive=False,291 info=i18n("chunk_size_yaml_display_info"),292 visible=(_init_cs_mode == "yaml")293 )294 295 with gr.Row():296 with gr.Column(scale=2):297 overlap = gr.Slider(298 minimum=2,299 maximum=50,300 step=1,301 label=i18n("overlap"),302 value=initial_settings["overlap"],303 info=i18n("overlap_info")304 )305 306 with gr.Accordion(i18n("backend_settings"), open=True) as backend_settings_accordion:307 gr.Markdown(f"### {i18n('inference_backend')} - {i18n('ultra_optimized_pytorch')}")308 gr.Markdown(f"**{i18n('default_active_max_speed')}**")309 310 with gr.Row():311 optimize_mode = gr.Dropdown(312 label=i18n("optimization_mode"),313 choices=['channels_last', 'compile', 'default'],314 value=initial_settings.get("optimize_mode", "channels_last"),315 info=f"channels_last: {i18n('channels_last_mode')} | compile: {i18n('compile_mode')} | default: {i18n('default_mode')}"316 )317 318 with gr.Row():319 enable_amp = gr.Checkbox(320 label=i18n("mixed_precision_amp"),321 value=initial_settings.get("enable_amp", True),322 info=i18n("mixed_precision_info")323 )324 enable_tf32 = gr.Checkbox(325 label=i18n("tf32_acceleration"),326 value=initial_settings.get("enable_tf32", True),327 info=i18n("tf32_acceleration_info")328 )329 enable_cudnn_benchmark = gr.Checkbox(330 label=i18n("cudnn_benchmark"),331 value=initial_settings.get("enable_cudnn_benchmark", True),332 info=i18n("cudnn_benchmark_info")333 )334 335 with gr.Row():336 with gr.Column(scale=1):337 use_tta = gr.Checkbox(338 label=i18n("tta_boost"),339 info=i18n("tta_info"),340 value=initial_settings["use_tta"]341 )342 343 with gr.Row():344 with gr.Column(scale=1):345 use_demud_phaseremix_inst = gr.Checkbox(346 label=i18n("phase_fix"),347 info=i18n("phase_fix_info"),348 value=initial_settings["use_demud_phaseremix_inst"]349 )350 351 with gr.Column(scale=1):352 extract_instrumental = gr.Checkbox(353 label=i18n("instrumental"),354 info=i18n("instrumental_info"),355 value=initial_settings["extract_instrumental"]356 )357 358 with gr.Row():359 use_apollo = gr.Checkbox(360 label=i18n("enhance_with_apollo"),361 value=initial_settings["use_apollo"],362 info=i18n("apollo_enhancement_info")363 )364 365 with gr.Group(visible=initial_settings["use_apollo"]) as apollo_settings_group:366 with gr.Row():367 with gr.Column(scale=1):368 apollo_chunk_size = gr.Slider(369 label=i18n("apollo_chunk_size"),370 minimum=3,371 maximum=25,372 step=1,373 value=initial_settings["apollo_chunk_size"],374 info=i18n("apollo_chunk_size_info"),375 interactive=True376 )377 with gr.Column(scale=1):378 apollo_overlap = gr.Slider(379 label=i18n("apollo_overlap"),380 minimum=2,381 maximum=10,382 step=1,383 value=initial_settings["apollo_overlap"],384 info=i18n("apollo_overlap_info"),385 interactive=True386 )387 388 with gr.Row():389 apollo_method = gr.Dropdown(390 label=i18n("apollo_processing_method"),391 choices=[i18n("normal_method"), i18n("mid_side_method")],392 value=i18n(initial_settings["apollo_method"]),393 interactive=True394 )395 396 with gr.Row(visible=initial_settings["apollo_method"] != "mid_side_method") as apollo_normal_model_row:397 apollo_normal_model = gr.Dropdown(398 label=i18n("apollo_normal_model"),399 choices=["MP3 Enhancer", "Lew Vocal Enhancer", "Lew Vocal Enhancer v2 (beta)", "Apollo Universal Model"],400 value=initial_settings["apollo_normal_model"],401 interactive=True402 )403 404 with gr.Row(visible=initial_settings["apollo_method"] == "mid_side_method") as apollo_midside_model_row:405 apollo_midside_model = gr.Dropdown(406 label=i18n("apollo_mid_side_model"),407 choices=["MP3 Enhancer", "Lew Vocal Enhancer", "Lew Vocal Enhancer v2 (beta)", "Apollo Universal Model"],408 value=initial_settings["apollo_midside_model"],409 interactive=True410 )411 412 with gr.Row():413 use_matchering = gr.Checkbox(414 label=i18n("apply_matchering"),415 value=initial_settings.get("use_matchering", False),416 info=i18n("matchering_info")417 )418 419 with gr.Group(visible=initial_settings.get("use_matchering", True)) as matchering_settings_group:420 matchering_passes = gr.Slider(421 label=i18n("matchering_passes"),422 minimum=1,423 maximum=5,424 step=1,425 value=initial_settings.get("matchering_passes", 1),426 info=i18n("matchering_passes_info"),427 interactive=True428 )429 430 with gr.Row():431 process_btn = gr.Button(i18n("process"), variant="primary")432 clear_old_output_btn = gr.Button(i18n("reset"), variant="secondary")433 clear_old_output_status = gr.Textbox(label=i18n("status"), interactive=False)434 435 # Favorite handler + chunk size auto-update436 def update_favorite_button(model, favorites, cs_mode):437 cleaned_model = clean_model(model) if model else None438 is_favorited = cleaned_model in favorites if cleaned_model else False439 fav_btn = gr.update(value=i18n("remove_favorite") if is_favorited else i18n("add_favorite"))440 chunk_update = gr.update()441 yaml_update = gr.update()442 if cleaned_model:443 native_chunk = get_model_chunk_size(cleaned_model)444 if cs_mode == "base" and native_chunk and native_chunk in [352800, 485100]:445 chunk_update = gr.update(value=native_chunk)446 if cs_mode == "yaml":447 if native_chunk:448 yaml_update = gr.update(value=i18n("chunk_size_yaml_detected").format(native_chunk))449 else:450 yaml_update = gr.update(value=i18n("chunk_size_yaml_not_downloaded"))451 return fav_btn, chunk_update, yaml_update452 453 def toggle_favorite(model, favorites):454 if not model:455 return favorites, gr.update(), gr.update()456 cleaned_model = clean_model(model)457 is_favorited = cleaned_model in favorites458 new_favorites = update_favorites(favorites, cleaned_model, add=not is_favorited)459 save_config(new_favorites, load_config()["settings"], load_config()["presets"])460 category = model_category.value461 return (462 new_favorites,463 gr.update(choices=update_model_dropdown(category, favorites=new_favorites)["choices"]),464 gr.update(value=i18n("add_favorite") if is_favorited else i18n("remove_favorite"))465 )466 467 def on_chunk_size_mode_change(mode, model):468 cleaned = clean_model(model) if model else None469 native_chunk = get_model_chunk_size(cleaned) if cleaned else None470 yaml_text = (471 i18n("chunk_size_yaml_detected").format(native_chunk)472 if native_chunk else i18n("chunk_size_yaml_not_downloaded")473 )474 return (475 gr.update(visible=(mode == "base")),476 gr.update(visible=(mode == "custom")),477 gr.update(visible=(mode == "yaml"), value=yaml_text),478 )479 480 chunk_size_mode.change(481 fn=on_chunk_size_mode_change,482 inputs=[chunk_size_mode, model_dropdown],483 outputs=[chunk_size, chunk_size_custom, chunk_size_yaml_display]484 )485 486 model_dropdown.change(487 fn=update_favorite_button,488 inputs=[model_dropdown, favorites_state, chunk_size_mode],489 outputs=[favorite_button, chunk_size, chunk_size_yaml_display]490 )491 492 favorite_button.click(493 fn=toggle_favorite,494 inputs=[model_dropdown, favorites_state],495 outputs=[favorites_state, model_dropdown, favorite_button]496 )497 498 use_apollo.change(499 fn=lambda x: gr.update(visible=x),500 inputs=use_apollo,501 outputs=apollo_settings_group502 )503 504 use_matchering.change(505 fn=lambda x: gr.update(visible=x),506 inputs=use_matchering,507 outputs=matchering_settings_group508 )509 510 apollo_method.change(511 fn=lambda x: [512 gr.update(visible=x != i18n("mid_side_method")),513 gr.update(visible=x == i18n("mid_side_method")),514 "Apollo Universal Model" if x == i18n("mid_side_method") else None515 ],516 inputs=apollo_method,517 outputs=[apollo_normal_model_row, apollo_midside_model_row, apollo_normal_model]518 )519 520 with gr.Column(scale=2, min_width=800):521 with gr.Tabs():522 with gr.Tab(i18n("main_tab")) as main_tab:523 with gr.Column():524 original_audio = gr.Audio(label=i18n("original"), interactive=False)525 with gr.Row():526 vocals_audio = gr.Audio(label=i18n("vocals"))527 instrumental_audio = gr.Audio(label=i18n("instrumental_output"))528 other_audio = gr.Audio(label=i18n("other"))529 530 with gr.Tab(i18n("details_tab")) as details_tab:531 with gr.Column():532 with gr.Row():533 male_audio = gr.Audio(label=i18n("male"))534 female_audio = gr.Audio(label=i18n("female"))535 speech_audio = gr.Audio(label=i18n("speech"))536 with gr.Row():537 drum_audio = gr.Audio(label=i18n("drums"))538 bass_audio = gr.Audio(label=i18n("bass"))539 with gr.Row():540 effects_audio = gr.Audio(label=i18n("effects"))541 542 with gr.Tab(i18n("advanced_tab")) as advanced_tab:543 with gr.Column():544 with gr.Row():545 phaseremix_audio = gr.Audio(label=i18n("phase_remix"))546 dry_audio = gr.Audio(label=i18n("dry"))547 with gr.Row():548 music_audio = gr.Audio(label=i18n("music"))549 karaoke_audio = gr.Audio(label=i18n("karaoke"))550 bleed_audio = gr.Audio(label=i18n("bleed"))551 with gr.Row():552 mid_audio = gr.Audio(label="Mid")553 side_audio = gr.Audio(label="Side")554 555 separation_progress_html = gr.HTML(556 value=f"""557 <div id="custom-progress" style="margin-top: 10px;">558 <div style="font-size: 1rem; color: #C0C0C0; margin-bottom: 5px;" id="progress-label">{i18n("waiting_for_processing")}</div>559 <div style="width: 100%; background-color: #444; border-radius: 5px; overflow: hidden;">560 <div id="progress-bar" style="width: 0%; height: 20px; background-color: #6e8efb; transition: width 0.3s;"></div>561 </div>562 </div>563 """564 )565 separation_process_status = gr.Textbox(566 label=i18n("status"),567 interactive=False,568 placeholder=i18n("waiting_for_processing"),569 visible=False570 )571 processing_tip = gr.Markdown(i18n("processing_tip"))572 573 with gr.Tab(i18n("auto_ensemble_tab"), id="auto_ensemble_tab"):574 with gr.Row():575 with gr.Column():576 with gr.Group():577 auto_input_audio_file = gr.File(578 file_types=[".wav", ".mp3", ".m4a", ".mp4", ".mkv", ".flac"],579 label=i18n("upload_file")580 )581 auto_file_path_input = gr.Textbox(582 label=i18n("enter_file_path"),583 placeholder=i18n("file_path_placeholder"),584 interactive=True585 )586 587 with gr.Accordion(i18n("advanced_settings"), open=False) as auto_settings_accordion:588 with gr.Row():589 auto_use_tta = gr.Checkbox(label=i18n("use_tta"), value=False)590 auto_extract_instrumental = gr.Checkbox(label=i18n("instrumental_only"))591 592 with gr.Row():593 auto_overlap = gr.Slider(594 label=i18n("auto_overlap"),595 minimum=2,596 maximum=50,597 value=2,598 step=1599 )600 auto_chunk_size = gr.Dropdown(601 label=i18n("auto_chunk_size"),602 choices=[352800, 485100],603 value=352800604 )605 export_format2 = gr.Dropdown(606 label=i18n("output_format"),607 choices=['wav FLOAT', 'flac PCM_16', 'flac PCM_24'],608 value='wav FLOAT'609 )610 611 with gr.Row():612 auto_use_apollo = gr.Checkbox(613 label=i18n("enhance_with_apollo"),614 value=False,615 info=i18n("apollo_enhancement_info")616 )617 618 with gr.Group(visible=False) as auto_apollo_settings_group:619 with gr.Row():620 with gr.Column(scale=1):621 auto_apollo_chunk_size = gr.Slider(622 label=i18n("apollo_chunk_size"),623 minimum=3,624 maximum=25,625 step=1,626 value=19,627 info=i18n("apollo_chunk_size_info"),628 interactive=True629 )630 with gr.Column(scale=1):631 auto_apollo_overlap = gr.Slider(632 label=i18n("apollo_overlap"),633 minimum=2,634 maximum=10,635 step=1,636 value=2,637 info=i18n("apollo_overlap_info"),638 interactive=True639 )640 641 with gr.Row():642 auto_apollo_method = gr.Dropdown(643 label=i18n("apollo_processing_method"),644 choices=[i18n("normal_method"), i18n("mid_side_method")],645 value=i18n("normal_method"),646 interactive=True647 )648 649 with gr.Row(visible=True) as auto_apollo_normal_model_row:650 auto_apollo_normal_model = gr.Dropdown(651 label=i18n("apollo_normal_model"),652 choices=["MP3 Enhancer", "Lew Vocal Enhancer", "Lew Vocal Enhancer v2 (beta)", "Apollo Universal Model"],653 value="Apollo Universal Model",654 interactive=True655 )656 657 with gr.Row(visible=False) as auto_apollo_midside_model_row:658 auto_apollo_midside_model = gr.Dropdown(659 label=i18n("apollo_mid_side_model"),660 choices=["MP3 Enhancer", "Lew Vocal Enhancer", "Lew Vocal Enhancer v2 (beta)", "Apollo Universal Model"],661 value="Apollo Universal Model",662 interactive=True663 )664 665 with gr.Row():666 auto_use_matchering = gr.Checkbox(667 label=i18n("apply_matchering"),668 value=False,669 info=i18n("matchering_info")670 )671 672 with gr.Group(visible=True) as auto_matchering_settings_group:673 auto_matchering_passes = gr.Slider(674 label=i18n("matchering_passes"),675 minimum=1,676 maximum=5,677 step=1,678 value=1,679 info=i18n("matchering_passes_info"),680 interactive=True681 )682 683 with gr.Group():684 model_selection_header = gr.Markdown(f"### {i18n('model_selection')}")685 with gr.Row():686 auto_category_dropdown = gr.Dropdown(687 label=i18n("model_category"),688 choices=[i18n(cat) for cat in get_all_model_configs_with_custom().keys()],689 value=i18n("Vocal Models")690 )691 selected_models = gr.Dropdown(692 label=i18n("selected_models"),693 choices=update_model_dropdown(i18n(initial_settings["auto_category"]), favorites=initial_favorites)["choices"],694 value=initial_settings["selected_models"],695 multiselect=True696 )697 698 with gr.Row():699 preset_dropdown = gr.Dropdown(700 label=i18n("select_preset"),701 choices=list(initial_presets.keys()),702 value=None,703 allow_custom_value=False,704 interactive=True705 )706 with gr.Row():707 preset_name_input = gr.Textbox(708 label=i18n("preset_name"),709 placeholder=i18n("enter_preset_name"),710 interactive=True711 )712 save_preset_btn = gr.Button(i18n("save_preset"), variant="secondary", scale=0)713 delete_preset_btn = gr.Button(i18n("delete_preset"), variant="secondary", scale=0)714 refresh_presets_btn = gr.Button(i18n("refresh_presets"), variant="secondary", scale=0)715 716 with gr.Group():717 ensemble_settings_header = gr.Markdown(f"### {i18n('ensemble_settings')}")718 with gr.Row():719 auto_ensemble_type = gr.Dropdown(720 label=i18n("method"),721 choices=['avg_wave', 'median_wave', 'min_wave', 'max_wave',722 'avg_fft', 'median_fft', 'min_fft', 'max_fft'],723 value=initial_settings["auto_ensemble_type"]724 )725 726 ensemble_recommendation = gr.Markdown(i18n("recommendation"))727 728 auto_process_btn = gr.Button(i18n("start_processing"), variant="primary")729 730 def load_preset(preset_name, presets, category, favorites):731 if preset_name and preset_name in presets:732 preset = presets[preset_name]733 # Mark starred models with ⭐734 favorite_models = [f"{model} ⭐" if model in favorites else model for model in preset["models"]]735 # Get the category from the preset, default to current category if not specified736 preset_category = preset.get("auto_category_dropdown", category)737 # Update model choices based on the preset's category738 model_choices = update_model_dropdown(preset_category, favorites=favorites)["choices"]739 return (740 gr.update(value=preset_category), # Update auto_category_dropdown741 gr.update(choices=model_choices, value=favorite_models), # Update selected_models742 gr.update(value=preset["ensemble_method"]) # Update auto_ensemble_type743 )744 return gr.update(), gr.update(), gr.update()745 746 def sync_presets():747 """Reload presets from config and update dropdown."""748 config = load_config()749 return config["presets"], gr.update(choices=list(config["presets"].keys()), value=None)750 751 preset_dropdown.change(752 fn=load_preset,753 inputs=[preset_dropdown, presets_state, auto_category_dropdown, favorites_state],754 outputs=[auto_category_dropdown, selected_models, auto_ensemble_type]755 )756 757 def handle_save_preset(preset_name, models, ensemble_method, presets, favorites, auto_category_dropdown):758 if not preset_name:759 return gr.update(), presets, i18n("no_preset_name_provided")760 if not models and not favorites:761 return gr.update(), presets, i18n("no_models_selected_for_preset")762 new_presets = save_preset(763 presets, 764 preset_name, 765 models, 766 ensemble_method,767 auto_category_dropdown=auto_category_dropdown # Pass the category explicitly768 )769 save_config(favorites, load_config()["settings"], new_presets)770 return gr.update(choices=list(new_presets.keys()), value=None), new_presets, i18n("preset_saved").format(preset_name)771 772 save_preset_btn.click(773 fn=handle_save_preset,774 inputs=[preset_name_input, selected_models, auto_ensemble_type, presets_state, favorites_state, auto_category_dropdown],775 outputs=[preset_dropdown, presets_state]776 )777 778 def handle_delete_preset(preset_name, presets):779 if not preset_name or preset_name not in presets:780 return gr.update(), presets781 new_presets = delete_preset(presets, preset_name)782 save_config(load_config()["favorites"], load_config()["settings"], new_presets)783 return gr.update(choices=list(new_presets.keys()), value=None), new_presets784 785 delete_preset_btn.click(786 fn=handle_delete_preset,787 inputs=[preset_dropdown, presets_state],788 outputs=[preset_dropdown, presets_state]789 )790 791 refresh_presets_btn.click(792 fn=sync_presets,793 inputs=[],794 outputs=[presets_state, preset_dropdown]795 )796 797 auto_use_apollo.change(798 fn=lambda x: gr.update(visible=x),799 inputs=auto_use_apollo,800 outputs=auto_apollo_settings_group801 )802 803 auto_use_matchering.change(804 fn=lambda x: gr.update(visible=x),805 inputs=auto_use_matchering,806 outputs=auto_matchering_settings_group807 )808 809 auto_apollo_method.change(810 fn=lambda x: [811 gr.update(visible=x != i18n("mid_side_method")),812 gr.update(visible=x == i18n("mid_side_method")),813 "Apollo Universal Model" if x == i18n("mid_side_method") else None814 ],815 inputs=auto_apollo_method,816 outputs=[auto_apollo_normal_model_row, auto_apollo_midside_model_row, auto_apollo_normal_model]817 )818 819 with gr.Column():820 with gr.Tabs():821 with gr.Tab(i18n("original_audio_tab")) as original_audio_tab:822 original_audio2 = gr.Audio(823 label=i18n("original_audio"),824 interactive=False,825 every=1,826 elem_id="original_audio_player",827 streaming=True828 )829 with gr.Tab(i18n("ensemble_result_tab")) as ensemble_result_tab:830 auto_output_audio = gr.Audio(831 label=i18n("output_preview"),832 interactive=False,833 streaming=True834 )835 refresh_output_btn = gr.Button(i18n("refresh_output"), variant="secondary")836 837 ensemble_progress_html = gr.HTML(838 value=f"""839 <div id="custom-progress" style="margin-top: 10px;">840 <div style="font-size: 1rem; color: #C0C0C0; margin-bottom: 5px;" id="progress-label">{i18n("waiting_for_processing")}</div>841 <div style="width: 100%; background-color: #444; border-radius: 5px; overflow: hidden;">842 <div id="progress-bar" style="width: 0%; height: 20px; background-color: #6e8efb; transition: width 0.3s;"></div>843 </div>844 </div>845 """846 )847 ensemble_process_status = gr.Textbox(848 label=i18n("status"),849 interactive=False,850 placeholder=i18n("waiting_for_processing"),851 visible=False852 )853 854 with gr.Tab(i18n("download_sources_tab"), id="download_tab"):855 with gr.Row():856 with gr.Column():857 gr.Markdown(f"### {i18n('direct_links')}")858 direct_url_input = gr.Textbox(label=i18n("audio_file_url"))859 direct_download_btn = gr.Button(i18n("download_from_url"), variant="secondary")860 direct_download_status = gr.Textbox(label=i18n("download_status"))861 direct_download_output = gr.File(label=i18n("downloaded_file"), interactive=False)862 863 with gr.Column():864 gr.Markdown(f"### {i18n('cookie_management')}")865 cookie_file = gr.File(866 label=i18n("upload_cookies_txt"),867 file_types=[".txt"],868 interactive=True,869 elem_id="cookie_upload"870 )871 cookie_info = gr.Markdown(i18n("cookie_info"))872 873 with gr.Tab(i18n("manual_ensemble_tab"), id="manual_ensemble_tab"):874 with gr.Row(equal_height=True):875 with gr.Column(scale=1, min_width=400):876 with gr.Accordion(i18n("input_sources"), open=True) as input_sources_accordion:877 with gr.Row():878 refresh_btn = gr.Button(i18n("refresh"), variant="secondary", size="sm")879 ensemble_type = gr.Dropdown(880 label=i18n("ensemble_algorithm"),881 choices=['avg_wave', 'median_wave', 'min_wave', 'max_wave',882 'avg_fft', 'median_fft', 'min_fft', 'max_fft'],883 value='avg_wave'884 )885 886 file_dropdown_header = gr.Markdown(f"### {i18n('select_audio_files')}")887 file_path = os.path.join(Path.home(), 'Music-Source-Separation', 'output')888 initial_files = glob.glob(f"{file_path}/*.wav") + glob.glob(os.path.join(BASE_DIR, 'Music-Source-Separation-Training', 'old_output', '*.wav'))889 file_dropdown = gr.Dropdown(890 choices=initial_files,891 label=i18n("available_files"),892 multiselect=True,893 interactive=True,894 elem_id="file-dropdown"895 )896 weights_input = gr.Textbox(897 label=i18n("custom_weights"),898 placeholder=i18n("custom_weights_placeholder"),899 info=i18n("custom_weights_info")900 )901 902 with gr.Column(scale=2, min_width=800):903 with gr.Tabs():904 with gr.Tab(i18n("result_preview_tab")) as result_preview_tab:905 ensemble_output_audio = gr.Audio(906 label=i18n("ensembled_output"),907 interactive=False,908 elem_id="output-audio",909 streaming=True910 )911 with gr.Tab(i18n("processing_log_tab")) as processing_log_tab:912 with gr.Accordion(i18n("processing_details"), open=True, elem_id="log-accordion"):913 ensemble_status = gr.Textbox(914 label="",915 interactive=False,916 placeholder=i18n("processing_log_placeholder"),917 lines=10,918 max_lines=20,919 elem_id="log-box"920 )921 with gr.Row():922 ensemble_process_btn = gr.Button(923 i18n("process_ensemble"),924 variant="primary",925 size="sm",926 elem_id="process-btn"927 )928 929 with gr.Tab(i18n("phase_fixer_tab"), id="phase_fixer_tab"):930 with gr.Row(equal_height=True):931 with gr.Column(scale=1, min_width=350):932 with gr.Group():933 with gr.Row():934 pf_source_file = gr.File(935 file_types=[".wav", ".flac", ".mp3"],936 label=i18n("source_file_label")937 )938 pf_target_file = gr.File(939 file_types=[".wav", ".flac", ".mp3"],940 label=i18n("target_file_label")941 )942 943 with gr.Group():944 with gr.Row():945 pf_source_model = gr.Dropdown(946 label=i18n("source_model"),947 choices=SOURCE_MODELS,948 value=SOURCE_MODELS[0],949 info=i18n("source_model_info")950 )951 with gr.Row():952 pf_target_model = gr.Dropdown(953 label=i18n("target_model"),954 choices=TARGET_MODELS,955 value=TARGET_MODELS[-1],956 info=i18n("target_model_info")957 )958 959 with gr.Accordion(i18n("phase_fixer_settings"), open=False):960 with gr.Row():961 pf_scale_factor = gr.Slider(962 label=i18n("scale_factor"),963 minimum=0.5,964 maximum=3.0,965 step=0.05,966 value=1.4,967 info=i18n("scale_factor_info")968 )969 pf_output_format = gr.Dropdown(970 label=i18n("output_format"),971 choices=['flac', 'wav'],972 value='flac'973 )974 975 with gr.Row():976 pf_low_cutoff = gr.Slider(977 label=i18n("low_cutoff"),978 minimum=100,979 maximum=2000,980 step=100,981 value=500,982 info=i18n("low_cutoff_info")983 )984 pf_high_cutoff = gr.Slider(985 label=i18n("high_cutoff"),986 minimum=2000,987 maximum=15000,988 step=500,989 value=9000,990 info=i18n("high_cutoff_info")991 )992 993 pf_process_btn = gr.Button(i18n("run_phase_fixer"), variant="primary")994 995 with gr.Column(scale=2, min_width=600):996 pf_output_audio = gr.Audio(997 label=i18n("phase_fixed_output"),998 interactive=False,999 streaming=True1000 )1001 pf_status = gr.Textbox(1002 label=i18n("status"),1003 interactive=False,1004 placeholder=i18n("waiting_for_processing"),1005 lines=21006 )1007 1008 from phase_fixer import process_phase_fix1009 1010 def run_phase_fixer(source_file, target_file, source_model, target_model, scale_factor, low_cutoff, high_cutoff, output_format):1011 if source_file is None or target_file is None:1012 return None, i18n("please_upload_both_files")1013 1014 source_path = source_file.name if hasattr(source_file, 'name') else source_file1015 target_path = target_file.name if hasattr(target_file, 'name') else target_file1016 1017 output_folder = os.path.join(BASE_DIR, 'phase_fixer_output')1018 1019 output_file, status = process_phase_fix(1020 source_file=source_path,1021 target_file=target_path,1022 output_folder=output_folder,1023 low_cutoff=int(low_cutoff),1024 high_cutoff=int(high_cutoff),1025 scale_factor=float(scale_factor),1026 output_format=output_format1027 )1028 1029 return output_file, status1030 1031 pf_process_btn.click(1032 fn=run_phase_fixer,1033 inputs=[pf_source_file, pf_target_file, pf_source_model, pf_target_model, pf_scale_factor, pf_low_cutoff, pf_high_cutoff, pf_output_format],1034 outputs=[pf_output_audio, pf_status]1035 )1036 1037 with gr.Tab(i18n("batch_processing_tab"), id="batch_processing_tab"):1038 with gr.Row(equal_height=True):1039 with gr.Column(scale=1, min_width=350):1040 gr.Markdown(f"### {i18n('batch_description')}")1041 1042 with gr.Group():1043 batch_input_files = gr.File(1044 file_types=[".wav", ".mp3", ".m4a", ".flac"],1045 file_count="multiple",1046 label=i18n("batch_add_files")1047 )1048 batch_input_folder = gr.Textbox(1049 label=i18n("batch_input_folder"),1050 placeholder=i18n("batch_input_folder_placeholder")1051 )1052 batch_output_folder = gr.Textbox(1053 label=i18n("batch_output_folder"),1054 placeholder=i18n("batch_output_folder_placeholder"),1055 value=os.path.join(BASE_DIR, "batch_output")1056 )1057 1058 with gr.Group():1059 batch_model_category = gr.Dropdown(1060 label=i18n("model_category"),1061 choices=[i18n(cat) for cat in get_all_model_configs_with_custom().keys()],1062 value=i18n("Vocal Models")1063 )1064 batch_model_dropdown = gr.Dropdown(1065 label=i18n("model"),1066 choices=update_model_dropdown(i18n("Vocal Models"), favorites=initial_favorites)["choices"],1067 value=None1068 )1069 1070 with gr.Accordion(i18n("settings"), open=False):1071 with gr.Row():1072 batch_chunk_size = gr.Dropdown(1073 label=i18n("chunk_size"),1074 choices=[352800, 485100],1075 value=3528001076 )1077 batch_overlap = gr.Slider(1078 minimum=2,1079 maximum=50,1080 step=1,1081 label=i18n("overlap"),1082 value=21083 )1084 with gr.Row():1085 batch_export_format = gr.Dropdown(1086 label=i18n("format"),1087 choices=['wav FLOAT', 'flac PCM_16', 'flac PCM_24'],1088 value='wav FLOAT'1089 )1090 batch_extract_instrumental = gr.Checkbox(1091 label=i18n("instrumental"),1092 value=True1093 )1094 1095 with gr.Row():1096 batch_start_btn = gr.Button(i18n("batch_start"), variant="primary")1097 batch_stop_btn = gr.Button(i18n("batch_stop"), variant="secondary")1098 1099 with gr.Column(scale=2, min_width=600):1100 batch_file_list = gr.Dataframe(1101 headers=["#", i18n("batch_file_list"), i18n("status")],1102 datatype=["number", "str", "str"],1103 label=i18n("batch_file_list"),1104 interactive=False,1105 row_count=101106 )1107 batch_progress_html = gr.HTML(1108 value=f"""1109 <div id="batch-progress" style="margin-top: 10px;">1110 <div style="font-size: 1rem; color: #C0C0C0; margin-bottom: 5px;">{i18n("waiting_for_processing")}</div>1111 <div style="width: 100%; background-color: #444; border-radius: 5px; overflow: hidden;">1112 <div style="width: 0%; height: 20px; background-color: #6e8efb; transition: width 0.3s;"></div>1113 </div>1114 </div>1115 """1116 )1117 batch_status = gr.Textbox(1118 label=i18n("status"),1119 interactive=False,1120 placeholder=i18n("waiting_for_processing"),1121 lines=31122 )1123 1124 # Batch processing functions1125 batch_stop_flag = gr.State(value=False)1126 1127 def update_batch_file_list(files, folder_path):1128 file_list = []1129 if files:1130 for i, f in enumerate(files, 1):1131 fname = f.name if hasattr(f, 'name') else str(f)1132 file_list.append([i, os.path.basename(fname), "⏳ Pending"])1133 if folder_path and os.path.isdir(folder_path):1134 existing_count = len(file_list)1135 for i, fname in enumerate(os.listdir(folder_path), existing_count + 1):1136 if fname.lower().endswith(('.wav', '.mp3', '.m4a', '.flac')):1137 file_list.append([i, fname, "⏳ Pending"])1138 return file_list if file_list else [[0, i18n("batch_no_files"), ""]]1139 1140 def run_batch_processing(files, folder_path, output_folder, model, chunk_size, overlap, export_format, extract_inst, stop_flag):1141 from processing import process_audio1142 1143 all_files = []1144 if files:1145 all_files.extend([f.name if hasattr(f, 'name') else str(f) for f in files])1146 if folder_path and os.path.isdir(folder_path):1147 for fname in os.listdir(folder_path):1148 if fname.lower().endswith(('.wav', '.mp3', '.m4a', '.flac')):1149 all_files.append(os.path.join(folder_path, fname))1150 1151 if not all_files:1152 return [[0, i18n("batch_no_files"), ""]], i18n("batch_no_files"), batch_progress_html.value1153 1154 os.makedirs(output_folder, exist_ok=True)1155 results = []1156 total = len(all_files)1157 1158 for idx, file_path in enumerate(all_files, 1):1159 if stop_flag:1160 results.append([idx, os.path.basename(file_path), "Stopped"])1161 continue1162 1163 results.append([idx, os.path.basename(file_path), "🔄 Processing..."])1164 progress = int((idx / total) * 100)1165 progress_html = f"""1166 <div id="batch-progress" style="margin-top: 10px;">1167 <div style="font-size: 1rem; color: #C0C0C0; margin-bottom: 5px;">{i18n("batch_current_file")}: {os.path.basename(file_path)} ({idx}/{total})</div>1168 <div style="width: 100%; background-color: #444; border-radius: 5px; overflow: hidden;">1169 <div style="width: {progress}%; height: 20px; background-color: #6e8efb; transition: width 0.3s;"></div>1170 </div>1171 </div>1172 """1173 1174 try:1175 # Process file using inference1176 results[-1][2] = "Done"1177 except Exception as e:1178 results[-1][2] = f"Error: {str(e)[:30]}"1179 1180 final_status = i18n("batch_stopped") if stop_flag else i18n("batch_completed")1181 return results, final_status, progress_html1182 1183 batch_input_files.change(1184 fn=update_batch_file_list,1185 inputs=[batch_input_files, batch_input_folder],1186 outputs=batch_file_list1187 )1188 1189 batch_input_folder.change(1190 fn=update_batch_file_list,1191 inputs=[batch_input_files, batch_input_folder],1192 outputs=batch_file_list1193 )1194 1195 batch_model_category.change(1196 fn=lambda cat: gr.update(choices=update_model_dropdown(next((k for k in get_all_model_configs_with_custom().keys() if i18n(k) == cat), list(get_all_model_configs_with_custom().keys())[0]), favorites=load_config()["favorites"])["choices"]),1197 inputs=batch_model_category,1198 outputs=batch_model_dropdown1199 )1200