modified: app.py

app.py

@@ -380,58 +380,58 @@ def generate_music(
                 instrumental_stem, _ = sf.read(vocal_path)
                 mix_stem = (vocal_stem + instrumental_stem) / 1
                 sf.write(recons_mix, mix_stem, sr)
-                return recons_mix
         except Exception as e:
             print(e)
+    return recons_mix
 
     # vocoder to upsample audios
-    vocal_decoder, inst_decoder = build_codec_model(config_path, vocal_decoder_path, inst_decoder_path)
-    vocoder_output_dir = os.path.join(output_dir, 'vocoder')
-    vocoder_stems_dir = os.path.join(vocoder_output_dir, 'stems')
-    vocoder_mix_dir = os.path.join(vocoder_output_dir, 'mix')
-    os.makedirs(vocoder_mix_dir, exist_ok=True)
-    os.makedirs(vocoder_stems_dir, exist_ok=True)
-    instrumental_output = None
-    vocal_output = None
-    for npy in stage1_output_set:
-        if 'instrumental' in npy:
-            # Process instrumental
-            instrumental_output = process_audio(
-                npy,
-                os.path.join(vocoder_stems_dir, 'instrumental.mp3'),
-                rescale,
-                argparse.Namespace(**locals()), # Convert local variables to argparse.Namespace
-                inst_decoder,
-                codec_model
-            )
-        else:
-            # Process vocal
-            vocal_output = process_audio(
-                npy,
-                os.path.join(vocoder_stems_dir, 'vocal.mp3'),
-                rescale,
-                 argparse.Namespace(**locals()), # Convert local variables to argparse.Namespace
-                vocal_decoder,
-                codec_model
-            )
-    # mix tracks
-    try:
-        mix_output = instrumental_output + vocal_output
-        vocoder_mix = os.path.join(vocoder_mix_dir, os.path.basename(recons_mix))
-        save_audio(mix_output, vocoder_mix, 44100, rescale)
-        print(f"Created mix: {vocoder_mix}")
-    except RuntimeError as e:
-        print(e)
-        print(f"mix {vocoder_mix} failed! inst: {instrumental_output.shape}, vocal: {vocal_output.shape}")
-
-    # Post process
-    replace_low_freq_with_energy_matched(
-        a_file=recons_mix,     # 16kHz
-        b_file=vocoder_mix,     # 48kHz
-        c_file=os.path.join(output_dir, os.path.basename(recons_mix)),
-        cutoff_freq=5500.0
-    )
-    print("All process Done")
+    # vocal_decoder, inst_decoder = build_codec_model(config_path, vocal_decoder_path, inst_decoder_path)
+    # vocoder_output_dir = os.path.join(output_dir, 'vocoder')
+    # vocoder_stems_dir = os.path.join(vocoder_output_dir, 'stems')
+    # vocoder_mix_dir = os.path.join(vocoder_output_dir, 'mix')
+    # os.makedirs(vocoder_mix_dir, exist_ok=True)
+    # os.makedirs(vocoder_stems_dir, exist_ok=True)
+    # instrumental_output = None
+    # vocal_output = None
+    # for npy in stage1_output_set:
+    #     if 'instrumental' in npy:
+    #         # Process instrumental
+    #         instrumental_output = process_audio(
+    #             npy,
+    #             os.path.join(vocoder_stems_dir, 'instrumental.mp3'),
+    #             rescale,
+    #             argparse.Namespace(**locals()), # Convert local variables to argparse.Namespace
+    #             inst_decoder,
+    #             codec_model
+    #         )
+    #     else:
+    #         # Process vocal
+    #         vocal_output = process_audio(
+    #             npy,
+    #             os.path.join(vocoder_stems_dir, 'vocal.mp3'),
+    #             rescale,
+    #              argparse.Namespace(**locals()), # Convert local variables to argparse.Namespace
+    #             vocal_decoder,
+    #             codec_model
+    #         )
+    # # mix tracks
+    # try:
+    #     mix_output = instrumental_output + vocal_output
+    #     vocoder_mix = os.path.join(vocoder_mix_dir, os.path.basename(recons_mix))
+    #     save_audio(mix_output, vocoder_mix, 44100, rescale)
+    #     print(f"Created mix: {vocoder_mix}")
+    # except RuntimeError as e:
+    #     print(e)
+    #     print(f"mix {vocoder_mix} failed! inst: {instrumental_output.shape}, vocal: {vocal_output.shape}")
+
+    # # Post process
+    # replace_low_freq_with_energy_matched(
+    #     a_file=recons_mix,     # 16kHz
+    #     b_file=vocoder_mix,     # 48kHz
+    #     c_file=os.path.join(output_dir, os.path.basename(recons_mix)),
+    #     cutoff_freq=5500.0
+    # )
+    # print("All process Done")
 
 
 @spaces.GPU(duration=120)