Update app.py

app.py

@@ -16,56 +16,60 @@ use_autocast = device == "cuda"
 #=========================#
 ROFORMER_MODELS = {
     # BS Roformer
-    'BS Roformer | Chorus Male-Female by Sucial': 'model_chorus_bs_roformer_ep_267_sdr_24.1275.ckpt',
-    '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',
     'BS-Roformer-Viperx-1297': 'model_bs_roformer_ep_317_sdr_12.9755.ckpt',
+    'BS-Roformer-De-Reverb': 'deverb_bs_roformer_8_384dim_10depth.ckpt',
+    'BS Roformer | Chorus Male-Female by Sucial': 'model_chorus_bs_roformer_ep_267_sdr_24.1275.ckpt',
 
     # MelBand Roformer
-    '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',
     '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',
-    'MelBand Roformer | De-Reverb by anvuew': 'dereverb_mel_band_roformer_anvuew_sdr_19.1729.ckpt',
-    'MelBand Roformer | De-Reverb Less Aggressive by anvuew': 'dereverb_mel_band_roformer_less_aggressive_anvuew_sdr_18.8050.ckpt',
-    'MelBand Roformer | De-Reverb-Echo by Sucial': 'dereverb-echo_mel_band_roformer_sdr_10.0169.ckpt',
-    'MelBand Roformer | De-Reverb-Echo V2 by Sucial': 'dereverb-echo_mel_band_roformer_sdr_13.4843_v2.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 | De-Reverb-Echo by Sucial': 'dereverb-echo_mel_band_roformer_sdr_10.0169.ckpt',
+    'MelBand Roformer | De-Reverb-Echo V2 by Sucial': 'dereverb-echo_mel_band_roformer_sdr_13.4843_v2.ckpt',
+    'MelBand Roformer | Aspiration Less Aggressive by Sucial': 'aspiration_mel_band_roformer_less_aggr_sdr_18.1201.ckpt',
+    'MelBand Roformer | Aspiration by Sucial': 'aspiration_mel_band_roformer_sdr_18.9845.ckpt',
+    'MelBand Roformer | De-Reverb Less Aggressive by anvuew': 'dereverb_mel_band_roformer_less_aggressive_anvuew_sdr_18.8050.ckpt',
+    'MelBand Roformer | De-Reverb by anvuew': 'dereverb_mel_band_roformer_anvuew_sdr_19.1729.ckpt',
     'MelBand Roformer | Vocals by Kimberley Jensen': 'vocals_mel_band_roformer.ckpt',
+    'Mel-Roformer-Denoise-Aufr33-Aggr': 'denoise_mel_band_roformer_aufr33_aggr_sdr_27.9768.ckpt',
+    'Mel-Roformer-Denoise-Aufr33': 'denoise_mel_band_roformer_aufr33_sdr_27.9959.ckpt',
+    'MelBand Roformer | Bleed Suppressor V1 by unwa-97chris': 'mel_band_roformer_bleed_suppressor_v1.ckpt',
 
     # MelBand Roformer Kim
+    'MelBand Roformer Kim | FT by unwa': 'mel_band_roformer_kim_ft_unwa.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 Kim | Big SYHFT V1 by SYH99999': 'MelBandRoformerBigSYHFTV1.ckpt',
-    'MelBand Roformer Kim | FT by unwa': 'mel_band_roformer_kim_ft_unwa.ckpt',
     'MelBand Roformer Kim | Inst V1 by Unwa': 'melband_roformer_inst_v1.ckpt',
     'MelBand Roformer Kim | Inst V1 (E) by Unwa': 'melband_roformer_inst_v1e.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',
+    '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 | SYHFT by SYH99999': 'MelBandRoformerSYHFT.ckpt',
+    'MelBand Roformer Kim | Big SYHFT V1 by SYH99999': 'MelBandRoformerBigSYHFTV1.ckpt',
 }
 #=========================#
 #      MDX23C Models      #
 #=========================#
 MDX23C_MODELS = {
+    'MDX23C DrumSep by aufr33-jarredou': 'MDX23C-DrumSep-aufr33-jarredou.ckpt',
+    'MDX23C De-Reverb by aufr33-jarredou': 'MDX23C-De-Reverb-aufr33-jarredou.ckpt',
     'MDX23C-InstVoc HQ': 'MDX23C-8KFFT-InstVoc_HQ.ckpt',
     'VIP | MDX23C-InstVoc HQ 2': 'MDX23C-8KFFT-InstVoc_HQ_2.ckpt',
     'VIP | MDX23C_D1581': 'MDX23C_D1581.ckpt',
-    'MDX23C De-Reverb by aufr33-jarredou': 'MDX23C-De-Reverb-aufr33-jarredou.ckpt',
-    'MDX23C DrumSep by aufr33-jarredou': 'MDX23C-DrumSep-aufr33-jarredou.ckpt',
 }
 #=========================#
 #     MDXN-NET Models     #
 #=========================#
 MDXNET_MODELS = {
+    'UVR-MDX-NET 1': 'UVR_MDXNET_1_9703.onnx',
+    'UVR-MDX-NET 2': 'UVR_MDXNET_2_9682.onnx',
+    'UVR-MDX-NET 3': 'UVR_MDXNET_3_9662.onnx',
+    'UVR_MDXNET_9482': 'UVR_MDXNET_9482.onnx',
     'UVR-MDX-NET Inst 1': 'UVR-MDX-NET-Inst_1.onnx',
     'UVR-MDX-NET Inst 2': 'UVR-MDX-NET-Inst_2.onnx',
     'UVR-MDX-NET Inst 3': 'UVR-MDX-NET-Inst_3.onnx',
@@ -74,25 +78,14 @@ MDXNET_MODELS = {
     'UVR-MDX-NET Inst HQ 3': 'UVR-MDX-NET-Inst_HQ_3.onnx',
     'UVR-MDX-NET Inst HQ 4': 'UVR-MDX-NET-Inst_HQ_4.onnx',
     'UVR-MDX-NET Inst HQ 5': 'UVR-MDX-NET-Inst_HQ_5.onnx',
-    'VIP | UVR-MDX-NET_Inst_82_beta': 'UVR-MDX-NET_Inst_82_beta.onnx',
-    'VIP | UVR-MDX-NET_Inst_90_beta': 'UVR-MDX-NET_Inst_90_beta.onnx',
-    'VIP | UVR-MDX-NET_Inst_187_beta': 'UVR-MDX-NET_Inst_187_beta.onnx',
-    'VIP | UVR-MDX-NET-Inst_full_292': 'UVR-MDX-NET-Inst_full_292.onnx',
-    'VIP | UVR-MDX-NET_Main_340': 'UVR-MDX-NET_Main_340.onnx',
-    'VIP | UVR-MDX-NET_Main_390': 'UVR-MDX-NET_Main_390.onnx',
-    'VIP | UVR-MDX-NET_Main_406': 'UVR-MDX-NET_Main_406.onnx',
-    'VIP | UVR-MDX-NET_Main_427': 'UVR-MDX-NET_Main_427.onnx',
-    'VIP | UVR-MDX-NET_Main_438': 'UVR-MDX-NET_Main_438.onnx',
-    'UVR-MDX-NET Crowd HQ 1 By Aufr33': 'UVR-MDX-NET-Crowd_HQ_1.onnx',
-    'UVR-MDX-NET Voc FT': 'UVR-MDX-NET-Voc_FT.onnx',
-    'UVR-MDX-NET 1': 'UVR_MDXNET_1_9703.onnx',
-    'UVR-MDX-NET 2': 'UVR_MDXNET_2_9682.onnx',
-    'UVR-MDX-NET 3': 'UVR_MDXNET_3_9662.onnx',
-    'UVR_MDXNET_9482': 'UVR_MDXNET_9482.onnx',
+    'UVR-MDX-NET Inst Main': 'UVR-MDX-NET-Inst_Main.onnx',
     'UVR-MDX-NET Karaoke': 'UVR_MDXNET_KARA.onnx',
     'UVR-MDX-NET Karaoke 2': 'UVR_MDXNET_KARA_2.onnx',
     'UVR-MDX-NET Main': 'UVR_MDXNET_Main.onnx',
-    'UVR-MDX-NET Inst Main': 'UVR-MDX-NET-Inst_Main.onnx',
+    'UVR-MDX-NET Voc FT': 'UVR-MDX-NET-Voc_FT.onnx',
+    'Kim Inst': 'Kim_Inst.onnx',
+    'Kim Vocal 1': 'Kim_Vocal_1.onnx',
+    'Kim Vocal 2': 'Kim_Vocal_2.onnx',
     'kuielab_a_bass': 'kuielab_a_bass.onnx',
     'kuielab_a_drums': 'kuielab_a_drums.onnx',
     'kuielab_a_other': 'kuielab_a_other.onnx',
@@ -101,10 +94,16 @@ MDXNET_MODELS = {
     'kuielab_b_drums': 'kuielab_b_drums.onnx',
     'kuielab_b_other': 'kuielab_b_other.onnx',
     'kuielab_b_vocals': 'kuielab_b_vocals.onnx',
-    'Kim Inst': 'Kim_Inst.onnx',
-    'Kim Vocal 1': 'Kim_Vocal_1.onnx',
-    'Kim Vocal 2': 'Kim_Vocal_2.onnx',
     'Reverb HQ By FoxJoy': 'Reverb_HQ_By_FoxJoy.onnx',
+    'VIP | UVR-MDX-NET_Inst_82_beta': 'UVR-MDX-NET_Inst_82_beta.onnx',
+    'VIP | UVR-MDX-NET_Inst_90_beta': 'UVR-MDX-NET_Inst_90_beta.onnx',
+    'VIP | UVR-MDX-NET_Inst_187_beta': 'UVR-MDX-NET_Inst_187_beta.onnx',
+    'VIP | UVR-MDX-NET-Inst_full_292': 'UVR-MDX-NET-Inst_full_292.onnx',
+    'VIP | UVR-MDX-NET_Main_340': 'UVR-MDX-NET_Main_340.onnx',
+    'VIP | UVR-MDX-NET_Main_390': 'UVR-MDX-NET_Main_390.onnx',
+    'VIP | UVR-MDX-NET_Main_406': 'UVR-MDX-NET_Main_406.onnx',
+    'VIP | UVR-MDX-NET_Main_427': 'UVR-MDX-NET_Main_427.onnx',
+    'VIP | UVR-MDX-NET_Main_438': 'UVR-MDX-NET_Main_438.onnx',
 }
 #========================#
 #     VR-ARCH Models     #
@@ -127,26 +126,26 @@ VR_ARCH_MODELS = {
     '15_SP-UVR-MID-44100-1': '15_SP-UVR-MID-44100-1.pth',
     '16_SP-UVR-MID-44100-2': '16_SP-UVR-MID-44100-2.pth',
     '17_HP-Wind_Inst-UVR': '17_HP-Wind_Inst-UVR.pth',
-    'UVR-De-Echo-Aggressive by FoxJoy': 'UVR-De-Echo-Aggressive.pth',
-    'UVR-De-Echo-Normal by FoxJoy': 'UVR-De-Echo-Normal.pth',
-    'UVR-DeEcho-DeReverb by FoxJoy': 'UVR-DeEcho-DeReverb.pth',
-    'UVR-DeNoise-Lite by FoxJoy': 'UVR-DeNoise-Lite.pth',
-    'UVR-DeNoise by FoxJoy': 'UVR-DeNoise.pth',
     'MGM_HIGHEND_v4': 'MGM_HIGHEND_v4.pth',
     'MGM_LOWEND_A_v4': 'MGM_LOWEND_A_v4.pth',
     'MGM_LOWEND_B_v4': 'MGM_LOWEND_B_v4.pth',
     'MGM_MAIN_v4': 'MGM_MAIN_v4.pth',
     'UVR-BVE-4B_SN-44100-1': 'UVR-BVE-4B_SN-44100-1.pth',
     'UVR-De-Reverb by aufr33-jarredou': 'UVR-De-Reverb-aufr33-jarredou.pth',
+    'UVR-De-Echo-Aggressive by FoxJoy': 'UVR-De-Echo-Aggressive.pth',
+    'UVR-De-Echo-Normal by FoxJoy': 'UVR-De-Echo-Normal.pth',
+    'UVR-DeEcho-DeReverb by FoxJoy': 'UVR-DeEcho-DeReverb.pth',
+    'UVR-DeNoise-Lite by FoxJoy': 'UVR-DeNoise-Lite.pth',
+    'UVR-DeNoise by FoxJoy': 'UVR-DeNoise.pth',
 }
 #=======================#
 #     DEMUCS Models     #
 #=======================#
 DEMUCS_MODELS = {
-    'hdemucs_mmi': 'hdemucs_mmi.yaml',
     'htdemucs': 'htdemucs.yaml',
     'htdemucs_6s': 'htdemucs_6s.yaml',
     'htdemucs_ft': 'htdemucs_ft.yaml',
+    'hdemucs_mmi': 'hdemucs_mmi.yaml',
 }
 
 def print_message(input_file, model_name):
@@ -408,7 +407,7 @@ with gr.Blocks(
     with gr.Tab("Roformer"):
         with gr.Group():
             with gr.Row():
-                roformer_model = gr.Dropdown(value="MelBand Roformer Kim | Inst V1 (E) - by Unwa", label="Select the Model", choices=list(ROFORMER_MODELS.keys()))
+                roformer_model = gr.Dropdown(value="MelBand Roformer Kim | Big Beta 5e FT by unwa", label="Select the Model", choices=list(ROFORMER_MODELS.keys()))
             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.")
@@ -425,7 +424,7 @@ with gr.Blocks(
     with gr.Tab("MDX23C"):
         with gr.Group():
             with gr.Row():
-                mdx23c_model = gr.Dropdown(label="Select the Model", choices=list(MDX23C_MODELS.keys()))
+                mdx23c_model = gr.Dropdown(value="MDX23C-InstVoc HQ", label="Select the Model", choices=list(MDX23C_MODELS.keys()))
             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.")
@@ -442,7 +441,7 @@ with gr.Blocks(
     with gr.Tab("MDX-NET"):
         with gr.Group():
             with gr.Row():
-                mdx_model = gr.Dropdown(label="Select the Model", choices=list(MDXNET_MODELS.keys()))
+                mdx_model = gr.Dropdown(value="UVR-MDX-NET Inst HQ 5", label="Select the Model", choices=list(MDXNET_MODELS.keys()))
             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.")
@@ -459,7 +458,7 @@ with gr.Blocks(
     with gr.Tab("VR ARCH"):
         with gr.Group():
             with gr.Row():
-                vr_model = gr.Dropdown(label="Select the Model", choices=list(VR_ARCH_MODELS.keys()))
+                vr_model = gr.Dropdown(value="1_HP-UVR", label="Select the Model", choices=list(VR_ARCH_MODELS.keys()))
             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=100, step=1, value=5, label="Agression", info="Intensity of primary stem extraction.")
@@ -478,7 +477,7 @@ with gr.Blocks(
     with gr.Tab("Demucs"):
         with gr.Group():
             with gr.Row():
-                demucs_model = gr.Dropdown(label="Select the Model", choices=list(DEMUCS_MODELS.keys()))
+                demucs_model = gr.Dropdown(value="htdemucs_6s", label="Select the Model", choices=list(DEMUCS_MODELS.keys()))
             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.")