Update app.py

app.py

@@ -1,9 +1,9 @@
 import os
 import torch
-import shutil
 import logging
+import yt_dlp
+import spaces
 import gradio as gr
-
 from audio_separator.separator import Separator
 
 device = "cuda" if torch.cuda.is_available() else "cpu"
@@ -12,80 +12,80 @@ use_autocast = device == "cuda"
 #=========================#
 #     Roformer Models     #
 #=========================#
-ROFORMER_MODELS = {
-    'BS-Roformer-De-Reverb': 'deverb_bs_roformer_8_384dim_10depth.ckpt',
-    'BS-Roformer-Viperx-1053': 'model_bs_roformer_ep_937_sdr_10.5309.ckpt',
-    'BS-Roformer-Viperx-1296': 'model_bs_roformer_ep_368_sdr_12.9628.ckpt',
+roformer_models = {
     'BS-Roformer-Viperx-1297': 'model_bs_roformer_ep_317_sdr_12.9755.ckpt',
+    'BS-Roformer-Viperx-1296': 'model_bs_roformer_ep_368_sdr_12.9628.ckpt',
+    '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',
+    'BS-Roformer-De-Reverb': 'deverb_bs_roformer_8_384dim_10depth.ckpt',
     'Mel-Roformer-Crowd-Aufr33-Viperx': 'mel_band_roformer_crowd_aufr33_viperx_sdr_8.7144.ckpt',
     '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',
+    'Mel-Roformer-Denoise-Aufr33-Aggr' : 'denoise_mel_band_roformer_aufr33_aggr_sdr_27.9768.ckpt',
     'Mel-Roformer-Karaoke-Aufr33-Viperx': 'mel_band_roformer_karaoke_aufr33_viperx_sdr_10.1956.ckpt',
-    'Mel-Roformer-Viperx-1143': 'model_mel_band_roformer_ep_3005_sdr_11.4360.ckpt',
-    'MelBand Roformer Kim | Inst V1 by Unwa': 'melband_roformer_inst_v1.ckpt',
-    'MelBand Roformer Kim | Inst V2 by Unwa': 'melband_roformer_inst_v2.ckpt',
-    'MelBand Roformer Kim | InstVoc Duality V1 by Unwa': 'melband_roformer_instvoc_duality_v1.ckpt',
-    'MelBand Roformer Kim | InstVoc Duality V2 by Unwa': 'melband_roformer_instvox_duality_v2.ckpt',
-    'Vocals Mel Band Roformer': 'vocals_mel_band_roformer.ckpt',
-    'Mel Band Roformer Bleed Suppressor V1': 'mel_band_roformer_bleed_suppressor_v1.ckpt',
-    'Mel Band Roformer SYHFT V2': 'MelBandRoformerSYHFTV2.ckpt',
-    'Mel Band Roformer SYHFT V2.5': 'MelBandRoformerSYHFTV2.5.ckpt',
+    'MelBand Roformer Kim | Inst V1 by Unwa' : 'melband_roformer_inst_v1.ckpt',
+    'MelBand Roformer Kim | Inst V2 by Unwa' : 'melband_roformer_inst_v2.ckpt',
+    'MelBand Roformer Kim | InstVoc Duality V1 by Unwa' : 'melband_roformer_instvoc_duality_v1.ckpt',
+    'MelBand Roformer Kim | InstVoc Duality V2 by Unwa' : 'melband_roformer_instvox_duality_v2.ckpt',
 }
+
 #=========================#
 #      MDX23C Models      #
 #=========================#
-MDX23C_MODELS = [
+mdx23c_models = [
+    'MDX23C_D1581.ckpt',
     'MDX23C-8KFFT-InstVoc_HQ.ckpt',
     'MDX23C-8KFFT-InstVoc_HQ_2.ckpt',
-    'MDX23C_D1581.ckpt',
 ]
+
 #=========================#
 #     MDXN-NET Models     #
 #=========================#
-MDXNET_MODELS = [
-    'UVR-MDX-NET-Crowd_HQ_1.onnx',
-    'UVR-MDX-NET-Inst_1.onnx',
-    'UVR-MDX-NET-Inst_2.onnx',
-    'UVR-MDX-NET-Inst_3.onnx',
-    'UVR-MDX-NET-Inst_HQ_1.onnx',
-    'UVR-MDX-NET-Inst_HQ_2.onnx',
-    'UVR-MDX-NET-Inst_HQ_3.onnx',
-    'UVR-MDX-NET-Inst_HQ_4.onnx',
-    'UVR-MDX-NET-Inst_HQ_5.onnx',
+mdxnet_models = [
     'UVR-MDX-NET-Inst_full_292.onnx',
-    'UVR-MDX-NET-Voc_FT.onnx',
+    'UVR-MDX-NET_Inst_187_beta.onnx',
     'UVR-MDX-NET_Inst_82_beta.onnx',
     'UVR-MDX-NET_Inst_90_beta.onnx',
-    'UVR-MDX-NET_Inst_187_beta.onnx',
     'UVR-MDX-NET_Main_340.onnx',
     'UVR-MDX-NET_Main_390.onnx',
     'UVR-MDX-NET_Main_406.onnx',
     'UVR-MDX-NET_Main_427.onnx',
     'UVR-MDX-NET_Main_438.onnx',
+    'UVR-MDX-NET-Inst_HQ_1.onnx',
+    'UVR-MDX-NET-Inst_HQ_2.onnx',
+    'UVR-MDX-NET-Inst_HQ_3.onnx',
+    'UVR-MDX-NET-Inst_HQ_4.onnx',
+    'UVR-MDX-NET-Inst_HQ_5.onnx',
+    'UVR_MDXNET_Main.onnx',
+    'UVR-MDX-NET-Inst_Main.onnx',
     'UVR_MDXNET_1_9703.onnx',
     'UVR_MDXNET_2_9682.onnx',
     'UVR_MDXNET_3_9662.onnx',
-    'UVR_MDXNET_9482.onnx',
+    'UVR-MDX-NET-Inst_1.onnx',
+    'UVR-MDX-NET-Inst_2.onnx',
+    'UVR-MDX-NET-Inst_3.onnx',
     'UVR_MDXNET_KARA.onnx',
     'UVR_MDXNET_KARA_2.onnx',
-    'UVR_MDXNET_Main.onnx',
+    'UVR_MDXNET_9482.onnx',
+    'UVR-MDX-NET-Voc_FT.onnx',
+    'Kim_Vocal_1.onnx',
+    'Kim_Vocal_2.onnx',
+    'Kim_Inst.onnx',
+    'Reverb_HQ_By_FoxJoy.onnx',
+    'UVR-MDX-NET_Crowd_HQ_1.onnx',
+    'kuielab_a_vocals.onnx',
+    'kuielab_a_other.onnx',
     'kuielab_a_bass.onnx',
     'kuielab_a_drums.onnx',
-    'kuielab_a_other.onnx',
-    'kuielab_a_vocals.onnx',
+    'kuielab_b_vocals.onnx',
+    'kuielab_b_other.onnx',
     'kuielab_b_bass.onnx',
     'kuielab_b_drums.onnx',
-    'kuielab_b_other.onnx',
-    'kuielab_b_vocals.onnx',
-    'Kim_Inst.onnx',
-    'Kim_Vocal_1.onnx',
-    'Kim_Vocal_2.onnx',
-    'Reverb_HQ_By_FoxJoy.onnx',
 ]
+
 #========================#
 #     VR-ARCH Models     #
 #========================#
-VR_ARCH_MODELS = [
+vrarch_models = [
     '1_HP-UVR.pth',
     '2_HP-UVR.pth',
     '3_HP-Vocal-UVR.pth',
@@ -103,168 +103,186 @@ VR_ARCH_MODELS = [
     '15_SP-UVR-MID-44100-1.pth',
     '16_SP-UVR-MID-44100-2.pth',
     '17_HP-Wind_Inst-UVR.pth',
-    'MGM_HIGHEND_v4.pth',
-    'MGM_LOWEND_A_v4.pth',
-    'MGM_LOWEND_B_v4.pth',
-    'MGM_MAIN_v4.pth',
-    'UVR-BVE-4B_SN-44100-1.pth',
-    'UVR-DeEcho-DeReverb.pth',
     'UVR-De-Echo-Aggressive.pth',
     'UVR-De-Echo-Normal.pth',
+    'UVR-DeEcho-DeReverb.pth',
     'UVR-DeNoise-Lite.pth',
     'UVR-DeNoise.pth',
+    'UVR-BVE-4B_SN-44100-1.pth',
+    'MGM_HIGHEND_v4.pth',
+    'MGM_LOWEND_A_v4.pth',
+    'MGM_LOWEND_B_v4.pth',
+    'MGM_MAIN_v4.pth',
 ]
+
 #=======================#
 #     DEMUCS Models     #
 #=======================#
-DEMUCS_MODELS = [
-    'hdemucs_mmi.yaml',
-    'htdemucs.yaml',
-    'htdemucs_6s.yaml',
+demucs_models = [
     'htdemucs_ft.yaml',
+    'htdemucs_6s.yaml',
+    'htdemucs.yaml',
+    'hdemucs_mmi.yaml',
+]
+
+output_format = [
+    'wav',
+    'flac',
+    'mp3',
+    'ogg',
+    'opus',
+    'm4a',
+    'aiff',
+    'ac3'
 ]
 
-def print_message(input_file, model_name):
-    """Prints information about the audio separation process."""
-    base_name = os.path.splitext(os.path.basename(input_file))[0]
-    print("\n")
-    print("🎵 Audio-Separator 🎵")
-    print("Input audio:", base_name)
-    print("Separation Model:", model_name)
-    print("Audio Separation Process...")
-
-def prepare_output_dir(input_file, output_dir):
-    """Create a directory for the output files and clean it if it already exists."""
-    base_name = os.path.splitext(os.path.basename(input_file))[0]
-    out_dir = os.path.join(output_dir, base_name)
+found_files = []
+logs = []
+out_dir = "./outputs"
+models_dir = "./models"
+extensions = (".wav", ".flac", ".mp3", ".ogg", ".opus", ".m4a", ".aiff", ".ac3")
+
+def download_audio(url, output_dir="ytdl"):
+
+    os.makedirs(output_dir, exist_ok=True)
+
+    ydl_opts = {
+        'format': 'bestaudio/best',
+        'postprocessors': [{
+            'key': 'FFmpegExtractAudio',
+            'preferredcodec': 'wav',
+            'preferredquality': '32',
+        }],
+        'outtmpl': os.path.join(output_dir, '%(title)s.%(ext)s'),
+        'postprocessor_args': [
+            '-acodec', 'pcm_f32le'
+        ],
+    }
+
     try:
-        if os.path.exists(out_dir):
-            shutil.rmtree(out_dir)
-        os.makedirs(out_dir)
+        with yt_dlp.YoutubeDL(ydl_opts) as ydl:
+            info = ydl.extract_info(url, download=False)
+            video_title = info['title']
+
+            ydl.download([url])
+
+            file_path = os.path.join(output_dir, f"{video_title}.wav")
+
+            if os.path.exists(file_path):
+                return os.path.abspath(file_path)
+            else:
+                raise Exception("Something went wrong")
+
     except Exception as e:
-        raise RuntimeError(f"Failed to prepare output directory {out_dir}: {e}")
-    return out_dir
+        raise Exception(f"Error extracting audio with yt-dlp: {str(e)}")
 
-def roformer_separator(audio, model_key, seg_size, override_seg_size, overlap, pitch_shift, model_dir, out_dir, out_format, norm_thresh, amp_thresh, batch_size, progress=gr.Progress()):
-    """Separate audio using Roformer model."""
+@spaces.GPU(duration=60)
+def roformer_separator(audio, model_key, out_format, segment_size, override_seg_size, overlap, batch_size, norm_thresh, amp_thresh, progress=gr.Progress(track_tqdm=True)):
     base_name = os.path.splitext(os.path.basename(audio))[0]
-    print_message(audio, model_key)
-    model = ROFORMER_MODELS[model_key]
+    roformer_model = roformer_models[model_key]
     try:
-        out_dir = prepare_output_dir(audio, out_dir)
         separator = Separator(
             log_level=logging.WARNING,
-            model_file_dir=model_dir,
+            model_file_dir=models_dir,
             output_dir=out_dir,
             output_format=out_format,
+            use_autocast=use_autocast,
             normalization_threshold=norm_thresh,
             amplification_threshold=amp_thresh,
-            use_autocast=use_autocast,
             mdxc_params={
-                "segment_size": seg_size,
+                "segment_size": segment_size,
                 "override_model_segment_size": override_seg_size,
                 "batch_size": batch_size,
                 "overlap": overlap,
-                "pitch_shift": pitch_shift,
             }
         )
+    
+        progress(0.2, desc="Loading model...")
+        separator.load_model(model_filename=roformer_model)
 
-        progress(0.2, desc="Model loaded...")
-        separator.load_model(model_filename=model)
-
-        progress(0.7, desc="Audio separated...")
+        progress(0.7, desc="Separating audio...")
         separation = separator.separate(audio, f"{base_name}_(Stem1)", f"{base_name}_(Stem2)")
-        print(f"Separation complete!\nResults: {', '.join(separation)}")
 
         stems = [os.path.join(out_dir, file_name) for file_name in separation]
         return stems[1], stems[0]
     except Exception as e:
         raise RuntimeError(f"Roformer separation failed: {e}") from e
 
-def mdx23c_separator(audio, model, seg_size, override_seg_size, overlap, pitch_shift, model_dir, out_dir, out_format, norm_thresh, amp_thresh, batch_size, progress=gr.Progress(track_tqdm=True)):
-    """Separate audio using MDX23C model."""
+@spaces.GPU(duration=60)
+def mdxc_separator(audio, model, out_format, segment_size, override_seg_size, overlap, batch_size, norm_thresh, amp_thresh, progress=gr.Progress(track_tqdm=True)):
     base_name = os.path.splitext(os.path.basename(audio))[0]
-    print_message(audio, model)
     try:
-        out_dir = prepare_output_dir(audio, out_dir)
         separator = Separator(
             log_level=logging.WARNING,
-            model_file_dir=model_dir,
+            model_file_dir=models_dir,
             output_dir=out_dir,
             output_format=out_format,
+            use_autocast=use_autocast,
             normalization_threshold=norm_thresh,
             amplification_threshold=amp_thresh,
-            use_autocast=use_autocast,
             mdxc_params={
-                "segment_size": seg_size,
+                "segment_size": segment_size,
                 "override_model_segment_size": override_seg_size,
                 "batch_size": batch_size,
                 "overlap": overlap,
-                "pitch_shift": pitch_shift,
             }
         )
 
-        progress(0.2, desc="Model loaded...")
+        progress(0.2, desc="Loading model...")
         separator.load_model(model_filename=model)
 
-        progress(0.7, desc="Audio separated...")
+        progress(0.7, desc="Separating audio...")
         separation = separator.separate(audio, f"{base_name}_(Stem1)", f"{base_name}_(Stem2)")
-        print(f"Separation complete!\nResults: {', '.join(separation)}")
 
         stems = [os.path.join(out_dir, file_name) for file_name in separation]
         return stems[1], stems[0]
     except Exception as e:
         raise RuntimeError(f"MDX23C separation failed: {e}") from e
 
-def mdx_separator(audio, model, hop_length, seg_size, overlap, denoise, model_dir, out_dir, out_format, norm_thresh, amp_thresh, batch_size, progress=gr.Progress()):
-    """Separate audio using MDX-NET model."""
+@spaces.GPU(duration=60)
+def mdxnet_separator(audio, model, out_format, hop_length, segment_size, denoise, overlap, batch_size, norm_thresh, amp_thresh, progress=gr.Progress(track_tqdm=True)):
     base_name = os.path.splitext(os.path.basename(audio))[0]
-    print_message(audio, model)
     try:
-        out_dir = prepare_output_dir(audio, out_dir)
         separator = Separator(
             log_level=logging.WARNING,
-            model_file_dir=model_dir,
+            model_file_dir=models_dir,
             output_dir=out_dir,
             output_format=out_format,
+            use_autocast=use_autocast,
             normalization_threshold=norm_thresh,
             amplification_threshold=amp_thresh,
-            use_autocast=use_autocast,
             mdx_params={
                 "hop_length": hop_length,
-                "segment_size": seg_size,
+                "segment_size": segment_size,
                 "overlap": overlap,
                 "batch_size": batch_size,
                 "enable_denoise": denoise,
             }
         )
 
-        progress(0.2, desc="Model loaded...")
+        progress(0.2, desc="Loading model...")
         separator.load_model(model_filename=model)
 
-        progress(0.7, desc="Audio separated...")
+        progress(0.7, desc="Separating audio...")
         separation = separator.separate(audio, f"{base_name}_(Stem1)", f"{base_name}_(Stem2)")
-        print(f"Separation complete!\nResults: {', '.join(separation)}")
 
         stems = [os.path.join(out_dir, file_name) for file_name in separation]
         return stems[0], stems[1]
     except Exception as e:
         raise RuntimeError(f"MDX-NET separation failed: {e}") from e
 
-def vr_separator(audio, model, window_size, aggression, tta, post_process, post_process_threshold, high_end_process, model_dir, out_dir, out_format, norm_thresh, amp_thresh, batch_size, progress=gr.Progress()):
-    """Separate audio using VR ARCH model."""
+@spaces.GPU(duration=60)
+def vrarch_separator(audio, model, out_format, window_size, aggression, tta, post_process, post_process_threshold, high_end_process, batch_size, norm_thresh, amp_thresh, progress=gr.Progress(track_tqdm=True)):
     base_name = os.path.splitext(os.path.basename(audio))[0]
-    print_message(audio, model)
     try:
-        out_dir = prepare_output_dir(audio, out_dir)
         separator = Separator(
             log_level=logging.WARNING,
-            model_file_dir=model_dir,
+            model_file_dir=models_dir,
             output_dir=out_dir,
             output_format=out_format,
+            use_autocast=use_autocast,
             normalization_threshold=norm_thresh,
             amplification_threshold=amp_thresh,
-            use_autocast=use_autocast,
             vr_params={
                 "batch_size": batch_size,
                 "window_size": window_size,
@@ -276,45 +294,43 @@ def vr_separator(audio, model, window_size, aggression, tta, post_process, post_
             }
         )
 
-        progress(0.2, desc="Model loaded...")
+        progress(0.2, desc="Loading model...")
         separator.load_model(model_filename=model)
 
-        progress(0.7, desc="Audio separated...")
+        progress(0.7, desc="Separating audio...")
         separation = separator.separate(audio, f"{base_name}_(Stem1)", f"{base_name}_(Stem2)")
-        print(f"Separation complete!\nResults: {', '.join(separation)}")
 
         stems = [os.path.join(out_dir, file_name) for file_name in separation]
         return stems[0], stems[1]
     except Exception as e:
         raise RuntimeError(f"VR ARCH separation failed: {e}") from e
 
-def demucs_separator(audio, model, seg_size, shifts, overlap, segments_enabled, model_dir, out_dir, out_format, norm_thresh, amp_thresh, progress=gr.Progress()):
-    """Separate audio using Demucs model."""
-    print_message(audio, model)
+@spaces.GPU(duration=60)
+def demucs_separator(audio, model, out_format, shifts, segment_size, segments_enabled, overlap, batch_size, norm_thresh, amp_thresh, progress=gr.Progress(track_tqdm=True)):
+    base_name = os.path.splitext(os.path.basename(audio))[0]
     try:
-        out_dir = prepare_output_dir(audio, out_dir)
         separator = Separator(
             log_level=logging.WARNING,
-            model_file_dir=model_dir,
+            model_file_dir=models_dir,
             output_dir=out_dir,
             output_format=out_format,
+            use_autocast=use_autocast,
             normalization_threshold=norm_thresh,
             amplification_threshold=amp_thresh,
-            use_autocast=use_autocast,
             demucs_params={
-                "segment_size": seg_size,
+                "batch_size": batch_size,
+                "segment_size": segment_size,
                 "shifts": shifts,
                 "overlap": overlap,
                 "segments_enabled": segments_enabled,
             }
         )
 
-        progress(0.2, desc="Model loaded...")
+        progress(0.2, desc="Loading model...")
         separator.load_model(model_filename=model)
 
-        progress(0.7, desc="Audio separated...")
+        progress(0.7, desc="Separating audio...")
         separation = separator.separate(audio)
-        print(f"Separation complete!\nResults: {', '.join(separation)}")
 
         stems = [os.path.join(out_dir, file_name) for file_name in separation]
         
@@ -331,231 +347,922 @@ def update_stems(model):
     else:
         return gr.update(visible=False)
 
+@spaces.GPU(duration=60)
+def roformer_batch(path_input, path_output, model_key, out_format, segment_size, override_seg_size, overlap, batch_size, norm_thresh, amp_thresh):
+    found_files.clear()
+    logs.clear()
+    roformer_model = roformer_models[model_key]
 
+    for audio_files in os.listdir(path_input):
+        if audio_files.endswith(extensions):
+            found_files.append(audio_files)
+    total_files = len(found_files)
 
+    if total_files == 0:
+        logs.append("No valid audio files.")
+        yield "\n".join(logs)
+    else:
+        logs.append(f"{total_files} audio files found")
+        found_files.sort()
 
+        for audio_files in found_files:
+            file_path = os.path.join(path_input, audio_files)
+            base_name = os.path.splitext(os.path.basename(file_path))[0]
+            try:
+                separator = Separator(
+                    log_level=logging.WARNING,
+                    model_file_dir=models_dir,
+                    output_dir=path_output,
+                    output_format=out_format,
+                    use_autocast=use_autocast,
+                    normalization_threshold=norm_thresh,
+                    amplification_threshold=amp_thresh,
+                    mdxc_params={
+                        "segment_size": segment_size,
+                        "override_model_segment_size": override_seg_size,
+                        "batch_size": batch_size,
+                        "overlap": overlap,
+                    }
+                )
 
+                logs.append("Loading model...")
+                yield "\n".join(logs)
+                separator.load_model(model_filename=roformer_model)
 
+                logs.append(f"Separating file: {audio_files}")
+                yield "\n".join(logs)
+                separator.separate(file_path, f"{base_name}_(Stem1)", f"{base_name}_(Stem2)")
+                logs.append(f"File: {audio_files} separated!")
+                yield "\n".join(logs)
+            except Exception as e:
+                raise RuntimeError(f"Roformer batch separation failed: {e}") from e
 
+@spaces.GPU(duration=60)
+def mdx23c_batch(path_input, path_output, model, out_format, segment_size, override_seg_size, overlap, batch_size, norm_thresh, amp_thresh):
+    found_files.clear()
+    logs.clear()
 
-with gr.Blocks(title="🎵 Audio-Separator 🎵",theme=gr.themes.Base()) as app:
-    gr.HTML("<h1> 🎵 Audio-Separator 🎵 </h1>")
-    
-    with gr.Tab("Roformer"):
-        with gr.Group():
-            with gr.Row():
-                roformer_model = gr.Dropdown(label="Select the Model", choices=list(ROFORMER_MODELS.keys()))
-            with gr.Row():
+    for audio_files in os.listdir(path_input):
+        if audio_files.endswith(extensions):
+            found_files.append(audio_files)
+    total_files = len(found_files)
+
+    if total_files == 0:
+        logs.append("No valid audio files.")
+        yield "\n".join(logs)
+    else:
+        logs.append(f"{total_files} audio files found")
+        found_files.sort()
+
+        for audio_files in found_files:
+            file_path = os.path.join(path_input, audio_files)
+            base_name = os.path.splitext(os.path.basename(file_path))[0]
+            try:
+                separator = Separator(
+                    log_level=logging.WARNING,
+                    model_file_dir=models_dir,
+                    output_dir=path_output,
+                    output_format=out_format,
+                    use_autocast=use_autocast,
+                    normalization_threshold=norm_thresh,
+                    amplification_threshold=amp_thresh,
+                    mdxc_params={
+                        "segment_size": segment_size,
+                        "override_model_segment_size": override_seg_size,
+                        "batch_size": batch_size,
+                        "overlap": overlap,
+                    }
+                )
+
+                logs.append("Loading model...")
+                yield "\n".join(logs)
+                separator.load_model(model_filename=model)
+
+                logs.append(f"Separating file: {audio_files}")
+                yield "\n".join(logs)
+                separator.separate(file_path, f"{base_name}_(Stem1)", f"{base_name}_(Stem2)")
+                logs.append(f"File: {audio_files} separated!")
+                yield "\n".join(logs)
+            except Exception as e:
+                raise RuntimeError(f"Roformer batch separation failed: {e}") from e
+
+@spaces.GPU(duration=60)
+def mdxnet_batch(path_input, path_output, model, out_format, hop_length, segment_size, denoise, overlap, batch_size, norm_thresh, amp_thresh):
+    found_files.clear()
+    logs.clear()
+
+    for audio_files in os.listdir(path_input):
+        if audio_files.endswith(extensions):
+            found_files.append(audio_files)
+    total_files = len(found_files)
+
+    if total_files == 0:
+        logs.append("No valid audio files.")
+        yield "\n".join(logs)
+    else:
+        logs.append(f"{total_files} audio files found")
+        found_files.sort()
+
+        for audio_files in found_files:
+            file_path = os.path.join(path_input, audio_files)
+            base_name = os.path.splitext(os.path.basename(file_path))[0]
+            try:
+                separator = Separator(
+                    log_level=logging.WARNING,
+                    model_file_dir=models_dir,
+                    output_dir=path_output,
+                    output_format=out_format,
+                    use_autocast=use_autocast,
+                    normalization_threshold=norm_thresh,
+                    amplification_threshold=amp_thresh,
+                    mdx_params={
+                        "hop_length": hop_length,
+                        "segment_size": segment_size,
+                        "overlap": overlap,
+                        "batch_size": batch_size,
+                        "enable_denoise": denoise,
+                    }
+                )
+
+                logs.append("Loading model...")
+                yield "\n".join(logs)
+                separator.load_model(model_filename=model)
+
+                logs.append(f"Separating file: {audio_files}")
+                yield "\n".join(logs)
+                separator.separate(file_path, f"{base_name}_(Stem1)", f"{base_name}_(Stem2)")
+                logs.append(f"File: {audio_files} separated!")
+                yield "\n".join(logs)
+            except Exception as e:
+                raise RuntimeError(f"Roformer batch separation failed: {e}") from e
+
+@spaces.GPU(duration=60)
+def vrarch_batch(path_input, path_output, model, out_format, window_size, aggression, tta, post_process, post_process_threshold, high_end_process, batch_size, norm_thresh, amp_thresh):
+    found_files.clear()
+    logs.clear()
+
+    for audio_files in os.listdir(path_input):
+        if audio_files.endswith(extensions):
+            found_files.append(audio_files)
+    total_files = len(found_files)
+
+    if total_files == 0:
+        logs.append("No valid audio files.")
+        yield "\n".join(logs)
+    else:
+        logs.append(f"{total_files} audio files found")
+        found_files.sort()
+
+        for audio_files in found_files:
+            file_path = os.path.join(path_input, audio_files)
+            base_name = os.path.splitext(os.path.basename(file_path))[0]
+            try:
+                separator = Separator(
+                    log_level=logging.WARNING,
+                    model_file_dir=models_dir,
+                    output_dir=path_output,
+                    output_format=out_format,
+                    use_autocast=use_autocast,
+                    normalization_threshold=norm_thresh,
+                    amplification_threshold=amp_thresh,
+                    vr_params={
+                        "batch_size": batch_size,
+                        "window_size": window_size,
+                        "aggression": aggression,
+                        "enable_tta": tta,
+                        "enable_post_process": post_process,
+                        "post_process_threshold": post_process_threshold,
+                        "high_end_process": high_end_process,
+                    }
+                )
+
+                logs.append("Loading model...")
+                yield "\n".join(logs)
+                separator.load_model(model_filename=model)
+
+                logs.append(f"Separating file: {audio_files}")
+                yield "\n".join(logs)
+                separator.separate(file_path, f"{base_name}_(Stem1)", f"{base_name}_(Stem2)")
+                logs.append(f"File: {audio_files} separated!")
+                yield "\n".join(logs)
+            except Exception as e:
+                raise RuntimeError(f"Roformer batch separation failed: {e}") from e
+
+@spaces.GPU(duration=60)
+def demucs_batch(path_input, path_output, model, out_format, shifts, segment_size, segments_enabled, overlap, batch_size, norm_thresh, amp_thresh):
+    found_files.clear()
+    logs.clear()
+
+    for audio_files in os.listdir(path_input):
+        if audio_files.endswith(extensions):
+            found_files.append(audio_files)
+    total_files = len(found_files)
+
+    if total_files == 0:
+        logs.append("No valid audio files.")
+        yield "\n".join(logs)
+    else:
+        logs.append(f"{total_files} audio files found")
+        found_files.sort()
+
+        for audio_files in found_files:
+            file_path = os.path.join(path_input, audio_files)
+            try:
+                separator = Separator(
+                    log_level=logging.WARNING,
+                    model_file_dir=models_dir,
+                    output_dir=path_output,
+                    output_format=out_format,
+                    use_autocast=use_autocast,
+                    normalization_threshold=norm_thresh,
+                    amplification_threshold=amp_thresh,
+                    demucs_params={
+                        "batch_size": batch_size,
+                        "segment_size": segment_size,
+                        "shifts": shifts,
+                        "overlap": overlap,
+                        "segments_enabled": segments_enabled,
+                    }
+                )
+
+                logs.append("Loading model...")
+                yield "\n".join(logs)
+                separator.load_model(model_filename=model)
+
+                logs.append(f"Separating file: {audio_files}")
+                yield "\n".join(logs)
+                separator.separate(file_path)
+                logs.append(f"File: {audio_files} separated!")
+                yield "\n".join(logs)
+            except Exception as e:
+                raise RuntimeError(f"Roformer batch separation failed: {e}") from e
+            
+with gr.Blocks(theme ="hev832/applio", title = "🎵 Audio Separator UI 🎵") as app:
+    with gr.Row():
+        gr.Markdown("<h1> 🎵 Audio Separator UI 🎵 </h1>")
+    with gr.Row():
+        with gr.Tabs():
+            with gr.TabItem("BS/Mel Roformer"):
                 with gr.Row():
-                    roformer_seg_size = gr.Slider(minimum=32, maximum=4000, step=32, value=256, label="Segment Size", info="Larger consumes more resources, but may give better results.")               
-                    roformer_override_seg_size = gr.Checkbox(value=False, label="Override segment size", info="Override model default segment size instead of using the model default value.")
+                    roformer_model = gr.Dropdown(
+                        label = "Select the model",
+                        choices = list(roformer_models.keys()),
+                        value = lambda : None,
+                        interactive = True
+                    )
+                    roformer_output_format = gr.Dropdown(
+                        label = "Select the output format",
+                        choices = output_format,
+                        value = lambda : None,
+                        interactive = True
+                    )
+                with gr.Accordion("Advanced settings"), open = False):
+                    with gr.Group():
+                        with gr.Row():
+                            roformer_segment_size = gr.Slider(
+                                label = "Segment size",
+                                info = "Larger consumes more resources, but may give better results",
+                                minimum = 32,
+                                maximum = 4000,
+                                step = 32,
+                                value = 256,
+                                interactive = True
+                            )
+                            roformer_override_segment_size = gr.Checkbox(
+                                label = "Override segment size",
+                                info = "Override model default segment size instead of using the model default value",
+                                value = False,
+                                interactive = True
+                            )
+                        with gr.Row():
+                            roformer_overlap = gr.Slider(
+                                label = "Overlap",
+                                info = "Amount of overlap between prediction windows",
+                                minimum = 2,
+                                maximum = 10,
+                                step = 1,
+                                value = 8,
+                                interactive = True
+                            )
+                            roformer_batch_size = gr.Slider(
+                                label = "Batch size",
+                                info = "Larger consumes more RAM but may process slightly faster",
+                                minimum = 1,
+                                maximum = 16,
+                                step = 1,
+                                value = 1,
+                                interactive = True
+                            )
+                        with gr.Row():
+                            roformer_normalization_threshold = gr.Slider(
+                                label = "Normalization threshold",
+                                info = "The threshold for audio normalization",
+                                minimum = 0.1,
+                                maximum = 1,
+                                step = 0.1,
+                                value = 0.1,
+                                interactive = True
+                            )
+                            roformer_amplification_threshold = gr.Slider(
+                                label = "Amplification threshold",
+                                info = "The threshold for audio amplification",
+                                minimum = 0.1,
+                                maximum = 1,
+                                step = 0.1,
+                                value = 0.1,
+                                interactive = True
+                            )
                 with gr.Row():
-                    roformer_overlap = gr.Slider(minimum=2, maximum=10, step=1, value=8, label="Overlap", info="Amount of overlap between prediction windows. Lower is better but slower.")              
-                    roformer_pitch_shift = gr.Slider(minimum=-12, maximum=12, step=1, value=0, label="Pitch shift", info="Shift audio pitch by a number of semitones while processing. may improve output for deep/high vocals.")
-        with gr.Row():
-            roformer_audio = gr.Audio(label="Input Audio", type="filepath")
-        with gr.Row():
-            roformer_button = gr.Button("Separate!", variant="primary")
-        with gr.Row():
-            roformer_stem1 = gr.Audio(label="Stem 1", type="filepath", interactive=False)
-            roformer_stem2 = gr.Audio(label="Stem 2", type="filepath", interactive=False)
-
-    with gr.Tab("MDX23C"):
-        with gr.Group():
-            with gr.Row():
-                mdx23c_model = gr.Dropdown(label="Select the Model", choices=MDX23C_MODELS)
-            with gr.Row():
-                mdx23c_seg_size = gr.Slider(minimum=32, maximum=4000, step=32, value=256, label="Segment Size", info="Larger consumes more resources, but may give better results.")
-                mdx23c_override_seg_size = gr.Checkbox(value=False, label="Override segment size", info="Override model default segment size instead of using the model default value.")
-            with gr.Row():
-                mdx23c_overlap = gr.Slider(minimum=2, maximum=50, step=1, value=8, label="Overlap", info="Amount of overlap between prediction windows. Higher is better but slower.")
-                mdx23c_pitch_shift = gr.Slider(minimum=-12, maximum=12, step=1, value=0, label="Pitch shift", info="Shift audio pitch by a number of semitones while processing. may improve output for deep/high vocals.")
-        with gr.Row():
-            mdx23c_audio = gr.Audio(label="Input Audio", type="filepath")
-        with gr.Row():
-            mdx23c_button = gr.Button("Separate!", variant="primary")
-        with gr.Row():
-            mdx23c_stem1 = gr.Audio(label="Stem 1", type="filepath", interactive=False)
-            mdx23c_stem2 = gr.Audio(label="Stem 2", type="filepath", interactive=False)
-
-    with gr.Tab("MDX-NET"):
-        with gr.Group():
-            with gr.Row():
-                mdx_model = gr.Dropdown(label="Select the Model", choices=MDXNET_MODELS)
-            with gr.Row():
-                mdx_hop_length = gr.Slider(minimum=32, maximum=2048, step=32, value=1024, label="Hop Length", info="Usually called stride in neural networks; only change if you know what you're doing.")
-                mdx_seg_size = gr.Slider(minimum=32, maximum=4000, step=32, value=256, label="Segment Size", info="Larger consumes more resources, but may give better results.")
-            with gr.Row():
-                mdx_overlap = gr.Slider(minimum=0.001, maximum=0.999, step=0.001, value=0.25, label="Overlap", info="Amount of overlap between prediction windows. Higher is better but slower.")
-                mdx_denoise = gr.Checkbox(value=False, label="Denoise", info="Enable denoising after separation.")
-        with gr.Row():
-            mdx_audio = gr.Audio(label="Input Audio", type="filepath")
-        with gr.Row():
-            mdx_button = gr.Button("Separate!", variant="primary")
-        with gr.Row():
-            mdx_stem1 = gr.Audio(label="Stem 1", type="filepath", interactive=False)
-            mdx_stem2 = gr.Audio(label="Stem 2", type="filepath", interactive=False)
-
-    with gr.Tab("VR ARCH"):
-        with gr.Group():
-            with gr.Row():
-                vr_model = gr.Dropdown(label="Select the Model", choices=VR_ARCH_MODELS)
-            with gr.Row():
-                vr_window_size = gr.Slider(minimum=320, maximum=1024, step=32, value=512, label="Window Size", info="Balance quality and speed. 1024 = fast but lower, 320 = slower but better quality.")
-                vr_aggression = gr.Slider(minimum=1, maximum=50, step=1, value=5, label="Agression", info="Intensity of primary stem extraction.")
-            with gr.Row():
-                vr_tta = gr.Checkbox(value=False, label="TTA", info="Enable Test-Time-Augmentation; slow but improves quality.")
-            with gr.Row():
-                vr_post_process = gr.Checkbox(value=False, label="Post Process", info="Identify leftover artifacts within vocal output; may improve separation for some songs.")
-                vr_post_process_threshold = gr.Slider(minimum=0.1, maximum=0.3, step=0.1, value=0.2, label="Post Process Threshold", info="Threshold for post-processing.")
-            vr_high_end_process = gr.Checkbox(value=False, label="High End Process", info="Mirror the missing frequency range of the output.")
-        with gr.Row():
-            vr_audio = gr.Audio(label="Input Audio", type="filepath")
-        with gr.Row():
-            vr_button = gr.Button("Separate!", variant="primary")
-        with gr.Row():
-            vr_stem1 = gr.Audio(label="Stem 1", type="filepath", interactive=False)
-            vr_stem2 = gr.Audio(label="Stem 2", type="filepath", interactive=False)
-
-    with gr.Tab("Demucs"):
-        with gr.Group():
-            with gr.Row():
-                demucs_model = gr.Dropdown(label="Select the Model", choices=DEMUCS_MODELS)
-            with gr.Row():
-                demucs_seg_size = gr.Slider(minimum=1, maximum=100, step=1, value=40, label="Segment Size", info="Size of segments into which the audio is split. Higher = slower but better quality.")
-                demucs_shifts = gr.Slider(minimum=0, maximum=20, step=1, value=2, label="Shifts", info="Number of predictions with random shifts, higher = slower but better quality.")
-                demucs_overlap = gr.Slider(minimum=0.001, maximum=0.999, step=0.001, value=0.25, label="Overlap", info="Overlap between prediction windows. Higher = slower but better quality.")
-                demucs_segments_enabled = gr.Checkbox(value=True, label="Segment-wise processing", info="Enable segment-wise processing.")
-        with gr.Row():
-            demucs_audio = gr.Audio(label="Input Audio", type="filepath")
-        with gr.Row():
-            demucs_button = gr.Button("Separate!", variant="primary")
-        with gr.Row():
-            demucs_stem1 = gr.Audio(label="Stem 1", type="filepath", interactive=False)
-            demucs_stem2 = gr.Audio(label="Stem 2", type="filepath", interactive=False)
-        with gr.Row():
-            demucs_stem3 = gr.Audio(label="Stem 3", type="filepath", interactive=False)
-            demucs_stem4 = gr.Audio(label="Stem 4", type="filepath", interactive=False)
-        with gr.Row(visible=False) as stem6:
-            demucs_stem5 = gr.Audio(label="Stem 5", type="filepath", interactive=False)
-            demucs_stem6 = gr.Audio(label="Stem 6", type="filepath", interactive=False)
-
-
-    with gr.Tab("General settings"):
-        with gr.Group():
-            model_file_dir = gr.Textbox(value="/tmp/audio-separator-models/", label="Directory to cache model files", info="The directory where model files are stored.", placeholder="/tmp/audio-separator-models/")
-            with gr.Row():
-                output_dir = gr.Textbox(value="output", label="File output directory", info="The directory where output files will be saved.", placeholder="output")
-                output_format = gr.Dropdown(value="wav", choices=["wav", "flac", "mp3"], label="Output Format", info="The format of the output audio file.")
-            with gr.Row():
-                norm_threshold = gr.Slider(minimum=0.1, maximum=1, step=0.1, value=0.9, label="Normalization threshold", info="The threshold for audio normalization.")
-                amp_threshold = gr.Slider(minimum=0.1, maximum=1, step=0.1, value=0.6, label="Amplification threshold", info="The threshold for audio amplification.")
-            with gr.Row():
-                batch_size = gr.Slider(minimum=1, maximum=16, step=1, value=1, label="Batch Size", info="Larger consumes more RAM but may process slightly faster.")
-    with gr.Tab("Credits"):
-        gr.Markdown("""
-        Politrees - gradio webui\n
-        theNeodev - mod the ui\n
-        nomadkaraoke - original project    
-        """)
+                    roformer_audio = gr.Audio(
+                        label = "Input audio",
+                        type = "filepath",
+                        interactive = True
+                    )
+                with gr.Accordion("Separation by link", open = False):
+                    with gr.Row():
+                        roformer_link = gr.Textbox(
+                            label = _("Link"),
+                            placeholder = "Paste the link here",
+                            interactive = True
+                        )
+                    with gr.Row():
+                        gr.Markdown("You can paste the link to the video/audio from many sites, check the complete list [here](https://github.com/yt-dlp/yt-dlp/blob/master/supportedsites.md)")
+                    with gr.Row():
+                        roformer_download_button = gr.Button(
+                            "Download!",
+                            variant = "primary"
+                        )
 
+                roformer_download_button.click(download_audio, [roformer_link], [roformer_audio])
+
+                
+
+                
+                with gr.Row():
+                    roformer_button = gr.Button(_("Separate!"), variant = "primary")
+                with gr.Row():
+                    roformer_stem1 = gr.Audio(
+                        show_download_button = True,
+                        interactive = False,
+                        label = _("Stem 1"),
+                        type = "filepath"
+                    )
+                    roformer_stem2 = gr.Audio(
+                        show_download_button = True,
+                        interactive = False,
+                        label = _("Stem 2"),
+                        type = "filepath"
+                    )
+
+                roformer_button.click(roformer_separator, [roformer_audio, roformer_model, roformer_output_format, roformer_segment_size, roformer_override_segment_size, roformer_overlap, roformer_batch_size, roformer_normalization_threshold, roformer_amplification_threshold], [roformer_stem1, roformer_stem2])
+
+            with gr.TabItem("MDX23C"):
+                with gr.Row():
+                    mdx23c_model = gr.Dropdown(
+                        label = _("Select the model"),
+                        choices = mdx23c_models,
+                        value = lambda : None,
+                        interactive = True
+                    )
+                    mdx23c_output_format = gr.Dropdown(
+                        label = _("Select the output format"),
+                        choices = output_format,
+                        value = lambda : None,
+                        interactive = True
+                    )
+                with gr.Accordion(_("Advanced settings"), open = False):
+                    with gr.Group():
+                        with gr.Row():
+                            mdx23c_segment_size = gr.Slider(
+                                minimum = 32,
+                                maximum = 4000,
+                                step = 32,
+                                label = _("Segment size"),
+                                info = _("Larger consumes more resources, but may give better results"),
+                                value = 256,
+                                interactive = True
+                            )
+                            mdx23c_override_segment_size = gr.Checkbox(
+                                label = _("Override segment size"),
+                                info = _("Override model default segment size instead of using the model default value"),
+                                value = False,
+                                interactive = True
+                            )
+                        with gr.Row():
+                            mdx23c_overlap = gr.Slider(
+                                minimum = 2,
+                                maximum = 50,
+                                step = 1,
+                                label = _("Overlap"),
+                                info = _("Amount of overlap between prediction windows"),
+                                value = 8,
+                                interactive = True
+                            )
+                            mdx23c_batch_size = gr.Slider(
+                                label = _("Batch size"),
+                                info = _("Larger consumes more RAM but may process slightly faster"),
+                                minimum = 1,
+                                maximum = 16,
+                                step = 1,
+                                value = 1,
+                                interactive = True
+                            )
+                        with gr.Row():
+                            mdx23c_normalization_threshold = gr.Slider(
+                                label = _("Normalization threshold"),
+                                info = _("The threshold for audio normalization"),
+                                minimum = 0.1,
+                                maximum = 1,
+                                step = 0.1,
+                                value = 0.1,
+                                interactive = True
+                            )
+                            mdx23c_amplification_threshold = gr.Slider(
+                                label = _("Amplification threshold"),
+                                info = _("The threshold for audio amplification"),
+                                minimum = 0.1,
+                                maximum = 1,
+                                step = 0.1,
+                                value = 0.1,
+                                interactive = True
+                            )
+                with gr.Row():
+                     mdx23c_audio = gr.Audio(
+                        label = _("Input audio"),
+                        type = "filepath",
+                        interactive = True
+                    )
+                with gr.Accordion(_("Separation by link"), open = False):
+                    with gr.Row():
+                        mdx23c_link = gr.Textbox(
+                            label = _("Link"),
+                            placeholder = _("Paste the link here"),
+                            interactive = True
+                        )
+                    with gr.Row():
+                        gr.Markdown(_("You can paste the link to the video/audio from many sites, check the complete list [here](https://github.com/yt-dlp/yt-dlp/blob/master/supportedsites.md)"))
+                    with gr.Row():
+                        mdx23c_download_button = gr.Button(
+                            _("Download!"),
+                            variant = "primary"
+                        )
+
+                mdx23c_download_button.click(download_audio, [mdx23c_link], [mdx23c_audio])
+
+                
+                with gr.Row():
+                    mdx23c_button = gr.Button(_("Separate!"), variant = "primary")
+                with gr.Row():
+                    mdx23c_stem1 = gr.Audio(
+                        show_download_button = True,
+                        interactive = False,
+                        label = _("Stem 1"),
+                        type = "filepath"
+                    )
+                    mdx23c_stem2 = gr.Audio(
+                        show_download_button = True,
+                        interactive = False,
+                        label = _("Stem 2"),
+                        type = "filepath"
+                    )
+
+                mdx23c_button.click(mdxc_separator, [mdx23c_audio, mdx23c_model, mdx23c_output_format, mdx23c_segment_size, mdx23c_override_segment_size, mdx23c_overlap, mdx23c_batch_size, mdx23c_normalization_threshold, mdx23c_amplification_threshold], [mdx23c_stem1, mdx23c_stem2])
+                
+            with gr.TabItem("MDX-NET"):
+                with gr.Row():
+                    mdxnet_model = gr.Dropdown(
+                        label = _("Select the model"),
+                        choices = mdxnet_models,
+                        value = lambda : None,
+                        interactive = True
+                    )
+                    mdxnet_output_format = gr.Dropdown(
+                        label = _("Select the output format"),
+                        choices = output_format,
+                        value = lambda : None,
+                        interactive = True
+                    )
+                with gr.Accordion(_("Advanced settings"), open = False):
+                    with gr.Group():
+                        with gr.Row():
+                            mdxnet_hop_length = gr.Slider(
+                                label = _("Hop length"),
+                                info = _("Usually called stride in neural networks; only change if you know what you're doing"),
+                                minimum = 32,
+                                maximum = 2048,
+                                step = 32,
+                                value = 1024,
+                                interactive = True
+                            )
+                            mdxnet_segment_size = gr.Slider(
+                                minimum = 32,
+                                maximum = 4000,
+                                step = 32,
+                                label = _("Segment size"),
+                                info = _("Larger consumes more resources, but may give better results"),
+                                value = 256,
+                                interactive = True
+                            )
+                            mdxnet_denoise = gr.Checkbox(
+                                label = _("Denoise"),
+                                info = _("Enable denoising during separation"),
+                                value = True,
+                                interactive = True
+                            )
+                        with gr.Row():
+                            mdxnet_overlap = gr.Slider(
+                                label = _("Overlap"),
+                                info = _("Amount of overlap between prediction windows"),
+                                minimum = 0.001,
+                                maximum = 0.999,
+                                step = 0.001,
+                                value = 0.25,
+                                interactive = True
+                            )
+                            mdxnet_batch_size = gr.Slider(
+                                label = _("Batch size"),
+                                info = _("Larger consumes more RAM but may process slightly faster"),
+                                minimum = 1,
+                                maximum = 16,
+                                step = 1,
+                                value = 1,
+                                interactive = True
+                            )
+                        with gr.Row():
+                            mdxnet_normalization_threshold = gr.Slider(
+                                label = _("Normalization threshold"),
+                                info = _("The threshold for audio normalization"),
+                                minimum = 0.1,
+                                maximum = 1,
+                                step = 0.1,
+                                value = 0.1,
+                                interactive = True
+                            )
+                            mdxnet_amplification_threshold = gr.Slider(
+                                label = _("Amplification threshold"),
+                                info = _("The threshold for audio amplification"),
+                                minimum = 0.1,
+                                maximum = 1,
+                                step = 0.1,
+                                value = 0.1,
+                                interactive = True
+                            )
+                with gr.Row():
+                    mdxnet_audio = gr.Audio(
+                        label = _("Input audio"),
+                        type = "filepath",
+                        interactive = True
+                    )
+                with gr.Accordion(_("Separation by link"), open = False):
+                    with gr.Row():
+                        mdxnet_link = gr.Textbox(
+                            label = _("Link"),
+                            placeholder = _("Paste the link here"),
+                            interactive = True
+                        )
+                    with gr.Row():
+                        gr.Markdown("You can paste the link to the video/audio from many sites, check the complete list [here](https://github.com/yt-dlp/yt-dlp/blob/master/supportedsites.md)")
+                    with gr.Row():
+                        mdxnet_download_button = gr.Button(
+                            "Download!",
+                            variant = "primary"
+                        )
+                
+                mdxnet_download_button.click(download_audio, [mdxnet_link], [mdxnet_audio])
+
+                
+                with gr.Row():
+                    mdxnet_button = gr.Button("Separate!", variant = "primary")
+                with gr.Row():
+                    mdxnet_stem1 = gr.Audio(
+                        show_download_button = True,
+                        interactive = False,
+                        label = "Stem 1",
+                        type = "filepath"
+                    )
+                    mdxnet_stem2 = gr.Audio(
+                        show_download_button = True,
+                        interactive = False,
+                        label = "Stem 2",
+                        type = "filepath"
+                    )
+
+                mdxnet_button.click(mdxnet_separator, [mdxnet_audio, mdxnet_model, mdxnet_output_format, mdxnet_hop_length, mdxnet_segment_size, mdxnet_denoise, mdxnet_overlap, mdxnet_batch_size, mdxnet_normalization_threshold, mdxnet_amplification_threshold], [mdxnet_stem1, mdxnet_stem2])
+
+            with gr.TabItem("VR ARCH"):
+                with gr.Row():
+                    vrarch_model = gr.Dropdown(
+                        label = "Select the model",
+                        choices = vrarch_models,
+                        value = lambda : None,
+                        interactive = True
+                    )
+                    vrarch_output_format = gr.Dropdown(
+                        label = "Select the output format",
+                        choices = output_format,
+                        value = lambda : None,
+                        interactive = True
+                    )
+                with gr.Accordion("Advanced settings", open = False):
+                    with gr.Group():
+                        with gr.Row():
+                            vrarch_window_size = gr.Slider(
+                                label = _("Window size"),
+                                info = _("Balance quality and speed. 1024 = fast but lower, 320 = slower but better quality"),
+                                minimum=320,
+                                maximum=1024,
+                                step=32,
+                                value = 512,
+                                interactive = True
+                            )
+                            vrarch_agression = gr.Slider(
+                                minimum = 1,
+                                maximum = 50,
+                                step = 1,
+                                label = "Agression",
+                                info = "Intensity of primary stem extraction",
+                                value = 5,
+                                interactive = True
+                            )
+                            vrarch_tta = gr.Checkbox(
+                                label = "TTA",
+                                info = "Enable Test-Time-Augmentation; slow but improves quality",
+                                value = True,
+                                visible = True,
+                                interactive = True
+                            )
+                        with gr.Row():
+                            vrarch_post_process = gr.Checkbox(
+                                label = "Post process",
+                                info = "Identify leftover artifacts within vocal output; may improve separation for some songs",
+                                value = False,
+                                visible = True,
+                                interactive = True
+                            )
+                            vrarch_post_process_threshold = gr.Slider(
+                                label = "Post process threshold",
+                                info = "Threshold for post-processing",
+                                minimum = 0.1,
+                                maximum = 0.3,
+                                step = 0.1,
+                                value = 0.2,
+                                interactive = True
+                            )
+                        with gr.Row():
+                            vrarch_high_end_process = gr.Checkbox(
+                                label = "High end process",
+                                info = "Mirror the missing frequency range of the output",
+                                value = False,
+                                visible = True,
+                                interactive = True,
+                            )
+                            vrarch_batch_size = gr.Slider(
+                                label = "Batch size",
+                                info = "Larger consumes more RAM but may process slightly faster",
+                                minimum = 1,
+                                maximum = 16,
+                                step = 1,
+                                value = 1,
+                                interactive = True
+                            )
+                        with gr.Row():
+                            vrarch_normalization_threshold = gr.Slider(
+                                label = "Normalization threshold",
+                                info = "The threshold for audio normalization",
+                                minimum = 0.1,
+                                maximum = 1,
+                                step = 0.1,
+                                value = 0.1,
+                                interactive = True
+                            )
+                            vrarch_amplification_threshold = gr.Slider(
+                                label = "Amplification threshold",
+                                info = "The threshold for audio amplification",
+                                minimum = 0.1,
+                                maximum = 1,
+                                step = 0.1,
+                                value = 0.1,
+                                interactive = True
+                            )
+                with gr.Row():
+                    vrarch_audio = gr.Audio(
+                        label = "Input audio",
+                        type = "filepath",
+                        interactive = True
+                    )
+                with gr.Accordion("Separation by link"), open = False):
+                    with gr.Row():
+                        vrarch_link = gr.Textbox(
+                            label = "Link",
+                            placeholder = _("Paste the link here"),
+                            interactive = True
+                        )
+                    with gr.Row():
+                        gr.Markdown("You can paste the link to the video/audio from many sites, check the complete list [here](https://github.com/yt-dlp/yt-dlp/blob/master/supportedsites.md)")
+                    with gr.Row():
+                        vrarch_download_button = gr.Button(
+                            "Download!",
+                            variant = "primary"
+                    )
+
+                vrarch_download_button.click(download_audio, [vrarch_link], [vrarch_audio])
             
-    demucs_model.change(update_stems, inputs=[demucs_model], outputs=stem6)
-
-    roformer_button.click(
-        roformer_separator,
-        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,
-        ],
-        outputs=[roformer_stem1, roformer_stem2],
-    )
-    mdx23c_button.click(
-        mdx23c_separator,
-        inputs=[
-            mdx23c_audio,
-            mdx23c_model,
-            mdx23c_seg_size,
-            mdx23c_override_seg_size,
-            mdx23c_overlap,
-            mdx23c_pitch_shift,
-            model_file_dir,
-            output_dir,
-            output_format,
-            norm_threshold,
-            amp_threshold,
-            batch_size,
-        ],
-        outputs=[mdx23c_stem1, mdx23c_stem2],
-    )
-    mdx_button.click(
-        mdx_separator,
-        inputs=[
-            mdx_audio,
-            mdx_model,
-            mdx_hop_length,
-            mdx_seg_size,
-            mdx_overlap,
-            mdx_denoise,
-            model_file_dir,
-            output_dir,
-            output_format,
-            norm_threshold,
-            amp_threshold,
-            batch_size,
-        ],
-        outputs=[mdx_stem1, mdx_stem2],
-    )
-    vr_button.click(
-        vr_separator,
-        inputs=[
-            vr_audio,
-            vr_model,
-            vr_window_size,
-            vr_aggression,
-            vr_tta,
-            vr_post_process,
-            vr_post_process_threshold,
-            vr_high_end_process,
-            model_file_dir,
-            output_dir,
-            output_format,
-            norm_threshold,
-            amp_threshold,
-            batch_size,
-        ],
-        outputs=[vr_stem1, vr_stem2],
-    )
-    demucs_button.click(
-        demucs_separator,
-        inputs=[
-            demucs_audio,
-            demucs_model,
-            demucs_seg_size,
-            demucs_shifts,
-            demucs_overlap,
-            demucs_segments_enabled,
-            model_file_dir,
-            output_dir,
-            output_format,
-            norm_threshold,
-            amp_threshold,
-        ],
-        outputs=[demucs_stem1, demucs_stem2, demucs_stem3, demucs_stem4, demucs_stem5, demucs_stem6],
-    )
+                
+                with gr.Row():
+                    vrarch_button = gr.Button("Separate!", variant = "primary")
+                with gr.Row():
+                    vrarch_stem1 = gr.Audio(
+                        show_download_button = True,
+                        interactive = False,
+                        type = "filepath",
+                        label = "Stem 1"
+                    )
+                    vrarch_stem2 = gr.Audio(
+                        show_download_button = True,
+                        interactive = False,
+                        type = "filepath",
+                        label = "Stem 2"
+                    )
 
-def main():
-    app.launch(share=True, debug=True)
+                vrarch_button.click(vrarch_separator, [vrarch_audio, vrarch_model, vrarch_output_format, vrarch_window_size, vrarch_agression, vrarch_tta, vrarch_post_process, vrarch_post_process_threshold, vrarch_high_end_process, vrarch_batch_size, vrarch_normalization_threshold, vrarch_amplification_threshold], [vrarch_stem1, vrarch_stem2])
+
+            with gr.TabItem("Demucs"):
+                with gr.Row():
+                    demucs_model = gr.Dropdown(
+                        label = "Select the model",
+                        choices = demucs_models,
+                        value = lambda : None,
+                        interactive = True
+                    )
+                    demucs_output_format = gr.Dropdown(
+                        label = "Select the output format",
+                        choices = output_format,
+                        value = lambda : None,
+                        interactive = True
+                    )
+                with gr.Accordion("Advanced settings", open = False):
+                    with gr.Group():
+                        with gr.Row():
+                            demucs_shifts = gr.Slider(
+                                label = "Shifts",
+                                info = "Number of predictions with random shifts, higher = slower but better quality",
+                                minimum = 1,
+                                maximum = 20,
+                                step = 1,
+                                value = 2,
+                                interactive = True
+                            )
+                            demucs_segment_size = gr.Slider(
+                                label = "Segment size",
+                                info = "Size of segments into which the audio is split. Higher = slower but better quality",
+                                minimum = 1,
+                                maximum = 100,
+                                step = 1,
+                                value = 40,
+                                interactive = True
+                            )
+                            demucs_segments_enabled = gr.Checkbox(
+                                label = "Segment-wise processing",
+                                info = "Enable segment-wise processing",
+                                value = True,
+                                interactive = True
+                            )
+                        with gr.Row():
+                            demucs_overlap = gr.Slider(
+                                label = "Overlap",
+                                info = "Overlap between prediction windows. Higher = slower but better quality",
+                                minimum=0.001,
+                                maximum=0.999,
+                                step=0.001,
+                                value = 0.25,
+                                interactive = True
+                            )
+                            demucs_batch_size = gr.Slider(
+                                label = "Batch size",
+                                info = "Larger consumes more RAM but may process slightly faster",
+                                minimum = 1,
+                                maximum = 16,
+                                step = 1,
+                                value = 1,
+                                interactive = True
+                            )
+                        with gr.Row():
+                            demucs_normalization_threshold = gr.Slider(
+                                label = "Normalization threshold",
+                                info = "The threshold for audio normalization",
+                                minimum = 0.1,
+                                maximum = 1,
+                                step = 0.1,
+                                value = 0.1,
+                                interactive = True
+                            )
+                            demucs_amplification_threshold = gr.Slider(
+                                label = "Amplification threshold",
+                                info = "The threshold for audio amplification",
+                                minimum = 0.1,
+                                maximum = 1,
+                                step = 0.1,
+                                value = 0.1,
+                                interactive = True
+                            )
+                with gr.Row():
+                    demucs_audio = gr.Audio(
+                        label = "Input audio",
+                        type = "filepath",
+                        interactive = True
+                    )
+                with gr.Accordion("Separation by link", open = False):
+                    with gr.Row():
+                        demucs_link = gr.Textbox(
+                            label = "Link",
+                            placeholder = "Paste the link here",
+                            interactive = True
+                    )
+                    with gr.Row():
+                        gr.Markdown("You can paste the link to the video/audio from many sites, check the complete list [here](https://github.com/yt-dlp/yt-dlp/blob/master/supportedsites.md)")
+                    with gr.Row():
+                        demucs_download_button = gr.Button(
+                            "Download!",
+                            variant = "primary"
+                        )
+
+                demucs_download_button.click(download_audio, [demucs_link], [demucs_audio])
+
+                
+                    with gr.Row():
+                        demucs_bath_button = gr.Button("Separate!", variant = "primary")
+                    with gr.Row():
+                        demucs_info = gr.Textbox(
+                            label = "Output information",
+                            interactive = False
+                        )
+
+                
+                with gr.Row():
+                    demucs_button = gr.Button("Separate!"), variant = "primary")
+                with gr.Row():
+                    demucs_stem1 = gr.Audio(
+                        show_download_button = True,
+                        interactive = False,
+                        type = "filepath",
+                        label = "Stem 1"
+                    )
+                    demucs_stem2 = gr.Audio(
+                        show_download_button = True,
+                        interactive = False,
+                        type = "filepath",
+                        label = "Stem 2"
+                    )
+                with gr.Row():
+                    demucs_stem3 = gr.Audio(
+                        show_download_button = True,
+                        interactive = False,
+                        type = "filepath",
+                        label = "Stem 3"
+                    )
+                    demucs_stem4 = gr.Audio(
+                        show_download_button = True,
+                        interactive = False,
+                        type = "filepath",
+                        label = "Stem 4"
+                    )
+                with gr.Row(visible=False) as stem6:
+                    demucs_stem5 = gr.Audio(
+                        show_download_button = True,
+                        interactive = False,
+                        type = "filepath",
+                        label = "Stem 5"
+                    )
+                    demucs_stem6 = gr.Audio(
+                        show_download_button = True,
+                        interactive = False,
+                        type = "filepath",
+                        label = "Stem 6"
+                    )
+
+                demucs_model.change(update_stems, inputs=[demucs_model], outputs=stem6)
+                
+                demucs_button.click(demucs_separator, [demucs_audio, demucs_model, demucs_output_format, demucs_shifts, demucs_segment_size, demucs_segments_enabled, demucs_overlap, demucs_batch_size, demucs_normalization_threshold, demucs_amplification_threshold], [demucs_stem1, demucs_stem2, demucs_stem3, demucs_stem4, demucs_stem5, demucs_stem6])
+
+            
+            with gr.TabItem("Credits"):
+                gr.Markdown(
+                    """
+                    audio separator UI created by **[Eddycrack 864] & [_noxty](https://huggingface.co/theNeofr).
+                    * python-audio-separator by [beveradb](https://github.com/beveradb).
+                    * Special thanks to [Ilaria](https://github.com/TheStingerX) for hosting this space and help.
+                    * Thanks to [Mikus](https://github.com/cappuch) for the help with the code.
+                    * Thanks to [Nick088](https://huggingface.co/Nick088) for the help to fix roformers.
+                    * Thanks to [yt_dlp](https://github.com/yt-dlp/yt-dlp) devs.
+                    * Separation by link source code and improvements by [Blane187](https://huggingface.co/Blane187).
+                    * Thanks to [ArisDev](https://github.com/aris-py) for porting UVR5 UI to Kaggle and improvements.
+                    * Thanks to [Bebra777228](https://github.com/Bebra777228)'s code for guiding me to improve my code.
+                    
+                    
+                    You can donate to the original UVR5 project here:
+                    [!["Buy Me A Coffee"](https://www.buymeacoffee.com/assets/img/custom_images/orange_img.png)](https://www.buymeacoffee.com/uvr5)
+                    """
+                )
 
-if __name__ == "__main__":
-    main()
+app.queue()
+app.launch(share=True, debug=True)