Update gui.py

gui.py

@@ -4,10 +4,11 @@ import torch
 import logging
 from yt_dlp import YoutubeDL
 import gradio as gr
+import argparse
+from audio_separator.separator import Separator
 import numpy as np
 import librosa
 import soundfile as sf
-from audio_separator.separator import Separator
 from ensemble import ensemble_files
 import shutil
 
@@ -22,8 +23,14 @@ logger = logging.getLogger(__name__)
 ROFORMER_MODELS = {
     "Vocals": {
         'MelBand Roformer | Big Beta 6X by unwa': 'melband_roformer_big_beta6x.ckpt',
+        'MelBand Roformer Kim | Big Beta 4 FT by unwa': 'melband_roformer_big_beta4.ckpt',
+        'MelBand Roformer Kim | Big Beta 5e FT by unwa': 'melband_roformer_big_beta5e.ckpt',
+        'MelBand Roformer | Big Beta 6 by unwa': 'melband_roformer_big_beta6.ckpt',
         'MelBand Roformer | Vocals by Kimberley Jensen': 'vocals_mel_band_roformer.ckpt',
         'MelBand Roformer Kim | FT 3 by unwa': 'mel_band_roformer_kim_ft3_unwa.ckpt',
+        'MelBand Roformer Kim | FT by unwa': 'mel_band_roformer_kim_ft_unwa.ckpt',
+        'MelBand Roformer Kim | FT 2 by unwa': 'mel_band_roformer_kim_ft2_unwa.ckpt',
+        'MelBand Roformer Kim | FT 2 Bleedless by unwa': 'mel_band_roformer_kim_ft2_bleedless_unwa.ckpt',
         'MelBand Roformer | Vocals by becruily': 'mel_band_roformer_vocals_becruily.ckpt',
         'MelBand Roformer | Vocals Fullness by Aname': 'mel_band_roformer_vocal_fullness_aname.ckpt',
         'BS Roformer | Vocals by Gabox': 'bs_roformer_vocals_gabox.ckpt',
@@ -81,6 +88,7 @@ ROFORMER_MODELS = {
         'Mel-Roformer-Denoise-Aufr33': 'denoise_mel_band_roformer_aufr33_sdr_27.9959.ckpt',
         'Mel-Roformer-Denoise-Aufr33-Aggr': 'denoise_mel_band_roformer_aufr33_aggr_sdr_27.9768.ckpt',
         'MelBand Roformer | Denoise-Debleed by Gabox': 'mel_band_roformer_denoise_debleed_gabox.ckpt',
+        'MelBand Roformer | Bleed Suppressor V1 by unwa-97chris': 'mel_band_roformer_bleed_suppressor_v1.ckpt',
     },
     "Karaoke": {
         'Mel-Roformer-Karaoke-Aufr33-Viperx': 'mel_band_roformer_karaoke_aufr33_viperx_sdr_10.1956.ckpt',
@@ -93,26 +101,19 @@ ROFORMER_MODELS = {
         'BS-Roformer-Viperx-1053': 'model_bs_roformer_ep_937_sdr_10.5309.ckpt',
         'Mel-Roformer-Viperx-1143': 'model_mel_band_roformer_ep_3005_sdr_11.4360.ckpt',
         'Mel-Roformer-Crowd-Aufr33-Viperx': 'mel_band_roformer_crowd_aufr33_viperx_sdr_8.7144.ckpt',
-        'MelBand Roformer Kim | FT by unwa': 'mel_band_roformer_kim_ft_unwa.ckpt',
-        'MelBand Roformer Kim | FT 2 by unwa': 'mel_band_roformer_kim_ft2_unwa.ckpt',
-        'MelBand Roformer Kim | FT 2 Bleedless by unwa': 'mel_band_roformer_kim_ft2_bleedless_unwa.ckpt',
         'MelBand Roformer Kim | SYHFT by SYH99999': 'MelBandRoformerSYHFT.ckpt',
         'MelBand Roformer Kim | SYHFT V2 by SYH99999': 'MelBandRoformerSYHFTV2.ckpt',
         'MelBand Roformer Kim | SYHFT V2.5 by SYH99999': 'MelBandRoformerSYHFTV2.5.ckpt',
         'MelBand Roformer Kim | SYHFT V3 by SYH99999': 'MelBandRoformerSYHFTV3Epsilon.ckpt',
         'MelBand Roformer Kim | Big SYHFT V1 by SYH99999': 'MelBandRoformerBigSYHFTV1.ckpt',
-        'MelBand Roformer Kim | Big Beta 4 FT by unwa': 'melband_roformer_big_beta4.ckpt',
-        'MelBand Roformer Kim | Big Beta 5e FT by unwa': 'melband_roformer_big_beta5e.ckpt',
-        'MelBand Roformer | Big Beta 6 by unwa': 'melband_roformer_big_beta6.ckpt',
         'MelBand Roformer | Aspiration by Sucial': 'aspiration_mel_band_roformer_sdr_18.9845.ckpt',
         'MelBand Roformer | Aspiration Less Aggressive by Sucial': 'aspiration_mel_band_roformer_less_aggr_sdr_18.1201.ckpt',
-        'MelBand Roformer | Bleed Suppressor V1 by unwa-97chris': 'mel_band_roformer_bleed_suppressor_v1.ckpt',
     }
 }
 
 OUTPUT_FORMATS = ['wav', 'flac', 'mp3', 'ogg', 'opus', 'm4a', 'aiff', 'ac3']
 
-# CSS (modified to remove logo.jpg reference)
+# CSS (değişmedi, aynı kalıyor)
 CSS = """
 /* Modern ve Etkileşimli Tema */
 #app-container {
@@ -130,7 +131,8 @@ CSS = """
     overflow: hidden;
 }
 body {
-    background: linear-gradient(135deg, #1a0b2e, #2e1a47);
+    background: url('/content/logo.jpg') no-repeat center center fixed;
+    background-size: cover;
     margin: 0;
     padding: 0;
     font-family: 'Roboto', sans-serif;
@@ -479,7 +481,7 @@ def create_interface():
             with gr.Tab("⚙️ Settings"):
                 with gr.Group(elem_classes="dubbing-theme"):
                     gr.Markdown("### General Settings")
-                    model_file_dir = gr.Textbox(value="models/", label="📂 Model Cache", placeholder="Path to model directory", interactive=True)
+                    model_file_dir = gr.Textbox(value="/tmp/audio-separator-models/", label="📂 Model Cache", placeholder="Path to model directory", interactive=True)
                     output_dir = gr.Textbox(value="output", label="📤 Output Directory", placeholder="Where to save results", interactive=True)
                     output_format = gr.Dropdown(value="wav", choices=OUTPUT_FORMATS, label="🎶 Output Format", interactive=True)
                     norm_threshold = gr.Slider(0.1, 1, value=0.9, step=0.1, label="🔊 Normalization Threshold", interactive=True)
@@ -493,7 +495,7 @@ def create_interface():
                     with gr.Row():
                         roformer_audio = gr.Audio(label="🎧 Upload Audio", type="filepath", interactive=True)
                         url_ro = gr.Textbox(label="🔗 Or Paste URL", placeholder="YouTube or audio URL", interactive=True)
-                        download_roformer = gr.Button("⬇️ Download", variant="secondary")
+                        download_roformer = gr.Button("⬇️ download", variant="secondary")
                     roformer_exclude_stems = gr.Textbox(label="🚫 Exclude Stems", placeholder="e.g., vocals, drums (comma-separated)", interactive=True)
                     with gr.Row():
                         roformer_category = gr.Dropdown(label="📚 Category", choices=list(ROFORMER_MODELS.keys()), value="General Purpose", interactive=True)
@@ -510,30 +512,30 @@ def create_interface():
                         roformer_stem2 = gr.Audio(label="🥁 Stem 2", type="filepath", interactive=False)
 
             # Auto Ensemble Sekmesi
-            #with gr.Tab("🎚️ Auto Ensemble"):
-            #    with gr.Group(elem_classes="dubbing-theme"):
-            #        gr.Markdown("### Ensemble Processing")
-            #        with gr.Row():
-            #            ensemble_audio = gr.Audio(label="🎧 Upload Audio", type="filepath", interactive=True)
-            #            url_ensemble = gr.Textbox(label="🔗 Or Paste URL", placeholder="YouTube or audio URL", interactive=True)
-            #            download_ensemble = gr.Button("⬇️ Download", variant="secondary")
-            #        ensemble_exclude_stems = gr.Textbox(label="🚫 Exclude Stems", placeholder="e.g., vocals, drums (comma-separated)", interactive=True)
-            #        with gr.Row():
-            #            ensemble_category = gr.Dropdown(label="📚 Category", choices=list(ROFORMER_MODELS.keys()), value="Instrumentals", interactive=True)
-            #            ensemble_models = gr.CheckboxGroup(label="🛠️ Models", choices=list(ROFORMER_MODELS["Instrumentals"].keys()), interactive=True)
-            #        with gr.Row():
-            #            ensemble_seg_size = gr.Slider(32, 4000, value=256, step=32, label="📏 Segment Size", interactive=True)
-            #            ensemble_overlap = gr.Slider(2, 10, value=8, step=1, label="🔄 Overlap", interactive=True)
-            #            ensemble_use_tta = gr.Checkbox(value=False, label="🔍 Use TTA", interactive=True)
-            #        ensemble_method = gr.Dropdown(label="⚙️ Ensemble Method", choices=['avg_wave', 'median_wave', 'max_wave', 'min_wave', 'avg_fft', 'median_fft', 'max_fft', 'min_fft'], value='avg_wave', interactive=True)
-            #        ensemble_weights = gr.Textbox(label="⚖️ Weights", placeholder="e.g., 1.0, 1.0 (comma-separated)", interactive=True)
-            #        ensemble_button = gr.Button("🎛️ Run Ensemble!", variant="primary")
-            #        ensemble_output = gr.Audio(label="🎶 Ensemble Result", type="filepath", interactive=False)
-            #        ensemble_status = gr.Textbox(label="📢 Status", interactive=False)
+            with gr.Tab("🎚️ Auto Ensemble"):
+                with gr.Group(elem_classes="dubbing-theme"):
+                    gr.Markdown("### Ensemble Processing")
+                    with gr.Row():
+                        ensemble_audio = gr.Audio(label="🎧 Upload Audio", type="filepath", interactive=True)
+                        url_ensemble = gr.Textbox(label="🔗 Or Paste URL", placeholder="YouTube or audio URL", interactive=True)
+                        download_ensemble = gr.Button("⬇️ download", variant="secondary")
+                    ensemble_exclude_stems = gr.Textbox(label="🚫 Exclude Stems", placeholder="e.g., vocals, drums (comma-separated)", interactive=True)
+                    with gr.Row():
+                        ensemble_category = gr.Dropdown(label="📚 Category", choices=list(ROFORMER_MODELS.keys()), value="Instrumentals", interactive=True)
+                        ensemble_models = gr.Dropdown(label="🛠️ Models", choices=list(ROFORMER_MODELS["Instrumentals"].keys()), multiselect=True, interactive=True)
+                    with gr.Row():
+                        ensemble_seg_size = gr.Slider(32, 4000, value=256, step=32, label="📏 Segment Size", interactive=True)
+                        ensemble_overlap = gr.Slider(2, 10, value=8, step=1, label="🔄 Overlap", interactive=True)
+                        ensemble_use_tta = gr.Checkbox(value=False, label="🔍 Use TTA", interactive=True)
+                    ensemble_method = gr.Dropdown(label="⚙️ Ensemble Method", choices=['avg_wave', 'median_wave', 'max_wave', 'min_wave', 'avg_fft', 'median_fft', 'max_fft', 'min_fft'], value='avg_wave', interactive=True)
+                    ensemble_weights = gr.Textbox(label="⚖️ Weights", placeholder="e.g., 1.0, 1.0 (comma-separated)", interactive=True)
+                    ensemble_button = gr.Button("🎛️ Run Ensemble!", variant="primary")
+                    ensemble_output = gr.Audio(label="🎶 Ensemble Result", type="filepath", interactive=False)
+                    ensemble_status = gr.Textbox(label="📢 Status", interactive=False)
 
         gr.HTML("<div class='footer'>Powered by Audio-Separator 🌟🎶 | Made with ❤️</div>")
 
-        # Event Handlers
+        # Event Handlers (Aynı kalıyor)
         roformer_category.change(update_roformer_models, inputs=[roformer_category], outputs=[roformer_model])
         download_roformer.click(fn=download_audio, inputs=[url_ro], outputs=[roformer_audio])
         roformer_button.click(
@@ -541,19 +543,26 @@ def create_interface():
             inputs=[roformer_audio, roformer_model, roformer_seg_size, roformer_override_seg_size, roformer_overlap, roformer_pitch_shift, model_file_dir, output_dir, output_format, norm_threshold, amp_threshold, batch_size, roformer_exclude_stems],
             outputs=[roformer_stem1, roformer_stem2]
         )
-      #  ensemble_category.change(update_ensemble_models, inputs=[ensemble_category], outputs=[ensemble_models])
-      #  download_ensemble.click(fn=download_audio, inputs=[url_ensemble], outputs=[ensemble_audio])
-      #  ensemble_button.click(
-      #      lambda *args: auto_ensemble_process(
-      #          *args[:-1],
-      #          weights=[float(w.strip()) for w in args[-1].split(',')] if args[-1] else None
-      #      ),
-      #      inputs=[ensemble_audio, ensemble_models, ensemble_seg_size, ensemble_overlap, output_format, ensemble_use_tta, model_file_dir, output_dir, norm_threshold, amp_threshold, batch_size, ensemble_method, ensemble_exclude_stems, ensemble_weights],
-      #      outputs=[ensemble_output, ensemble_status]
-      #  )
+        ensemble_category.change(update_ensemble_models, inputs=[ensemble_category], outputs=[ensemble_models])
+        download_ensemble.click(fn=download_audio, inputs=[url_ensemble], outputs=[ensemble_audio])
+        ensemble_button.click(
+            lambda *args: auto_ensemble_process(
+                *args[:-1],
+                weights=[float(w.strip()) for w in args[-1].split(',')] if args[-1] else None
+            ),
+            inputs=[ensemble_audio, ensemble_models, ensemble_seg_size, ensemble_overlap, output_format, ensemble_use_tta, model_file_dir, output_dir, norm_threshold, amp_threshold, batch_size, ensemble_method, ensemble_exclude_stems, ensemble_weights],
+            outputs=[ensemble_output, ensemble_status]
+        )
     
     return app
 
 if __name__ == "__main__":
+    parser = argparse.ArgumentParser(description="Music Source Separation Web UI")
+    parser.add_argument("--port", type=int, default=7860, help="Port to run the UI on")
+    parser.add_argument("--ngrok-token", type=str, default=None, help="Ngrok token for tunneling")
+    args = parser.parse_args()
+
     app = create_interface()
-    app.launch()
+    app.launch(server_name="0.0.0.0", server_port=args.port, share=True)
+       
+    app.close()