updated with dropdown

app.py

@@ -1,25 +1,83 @@
+# import os
+# import gradio as gr
+# from scipy.io.wavfile import write
+# import subprocess
+# import torch
+
+# from audio_separator import Separator  # Ensure this is correctly implemented
+
+# def inference(audio):
+#     os.makedirs("out", exist_ok=True)
+#     audio_path = 'test.wav'
+#     write(audio_path, audio[0], audio[1])
+#     device = 'cuda' if torch.cuda.is_available() else 'cpu'
+#     if device=='cuda':
+#         use_cuda=True
+#         print(f"Using device: {device}")
+#     else:
+#         use_cuda=False
+#         print(f"Using device: {device}")
+#     try:
+        
+#         # Using subprocess.run for better control
+#         command = f"python3 -m demucs.separate -n htdemucs_6s -d {device} {audio_path} -o out"
+#         process = subprocess.run(command, shell=True, check=True, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
+#         print("Demucs script output:", process.stdout.decode())
+#     except subprocess.CalledProcessError as e:
+#         print("Error in Demucs script:", e.stderr.decode())
+#         return None
+
+#     try:
+#         # Separating the stems using your custom separator
+#         separator = Separator("./out/htdemucs_6s/test/vocals.wav", model_name='UVR_MDXNET_KARA_2', use_cuda=use_cuda, output_format='mp3')
+#         primary_stem_path, secondary_stem_path = separator.separate()
+#     except Exception as e:
+#         print("Error in custom separation:", str(e))
+#         return None
+
+#     # Collecting all file paths
+#     files = [f"./out/htdemucs_6s/test/{stem}.wav" for stem in ["vocals", "bass", "drums", "other", "piano", "guitar"]]
+#     files.extend([secondary_stem_path,primary_stem_path ])
+
+#     # Check if files exist
+#     existing_files = [file for file in files if os.path.isfile(file)]
+#     if not existing_files:
+#         print("No files were created.")
+#         return None
+
+#     return existing_files
+
+# # Gradio Interface
+# title = "Source Separation Demo"
+# description = "Music Source Separation in the Waveform Domain. To use it, simply upload your audio."
+# gr.Interface(
+#     inference, 
+#     gr.components.Audio(type="numpy", label="Input"), 
+#     [gr.components.Audio(type="filepath", label=stem) for stem in ["Full Vocals","Bass", "Drums", "Other", "Piano", "Guitar", "Lead Vocals", "Backing Vocals" ]],
+#     title=title,
+#     description=description,
+# ).launch()
+
+
 import os
 import gradio as gr
 from scipy.io.wavfile import write
 import subprocess
 import torch
 
-from audio_separator import Separator  # Ensure this is correctly implemented
+from audio_separator import Separator
 
 def inference(audio):
     os.makedirs("out", exist_ok=True)
     audio_path = 'test.wav'
     write(audio_path, audio[0], audio[1])
+
+    # Check for CUDA availability
     device = 'cuda' if torch.cuda.is_available() else 'cpu'
-    if device=='cuda':
-        use_cuda=True
-        print(f"Using device: {device}")
-    else:
-        use_cuda=False
-        print(f"Using device: {device}")
+    print(f"Using device: {device}")
+
+    # Run Demucs script
     try:
-        
-        # Using subprocess.run for better control
         command = f"python3 -m demucs.separate -n htdemucs_6s -d {device} {audio_path} -o out"
         process = subprocess.run(command, shell=True, check=True, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
         print("Demucs script output:", process.stdout.decode())
@@ -27,8 +85,9 @@ def inference(audio):
         print("Error in Demucs script:", e.stderr.decode())
         return None
 
+    # Run custom separator
+    use_cuda = device == 'cuda'
     try:
-        # Separating the stems using your custom separator
         separator = Separator("./out/htdemucs_6s/test/vocals.wav", model_name='UVR_MDXNET_KARA_2', use_cuda=use_cuda, output_format='mp3')
         primary_stem_path, secondary_stem_path = separator.separate()
     except Exception as e:
@@ -36,24 +95,39 @@ def inference(audio):
         return None
 
     # Collecting all file paths
-    files = [f"./out/htdemucs_6s/test/{stem}.wav" for stem in ["vocals", "bass", "drums", "other", "piano", "guitar"]]
-    files.extend([secondary_stem_path,primary_stem_path ])
+    stem_names = ["vocals", "bass", "drums", "other", "piano", "guitar", "lead_vocals", "backing_vocals"]
+    files = {stem: f"./out/htdemucs_6s/test/{stem}.wav" for stem in stem_names[:6]}
+    files["lead_vocals"] = primary_stem_path
+    files["backing_vocals"] = secondary_stem_path
 
-    # Check if files exist
-    existing_files = [file for file in files if os.path.isfile(file)]
+    # Check if files exist and return dictionary
+    existing_files = {stem: file for stem, file in files.items() if os.path.isfile(file)}
     if not existing_files:
         print("No files were created.")
         return None
 
     return existing_files
 
+# Function to return selected audio
+def get_selected_audio(stems, selected_stem):
+    return stems.get(selected_stem, None)
+
 # Gradio Interface
 title = "Source Separation Demo"
 description = "Music Source Separation in the Waveform Domain. To use it, simply upload your audio."
+
+audio_input = gr.components.Audio(type="numpy", label="Input")
+dropdown = gr.components.Dropdown(label="Select Stem", choices=["vocals", "bass", "drums", "other", "piano", "guitar", "lead_vocals", "backing_vocals"])
+audio_output = gr.components.Audio(type="filepath", label="Output")
+
 gr.Interface(
-    inference, 
-    gr.components.Audio(type="numpy", label="Input"), 
-    [gr.components.Audio(type="filepath", label=stem) for stem in ["Full Vocals","Bass", "Drums", "Other", "Piano", "Guitar", "Lead Vocals", "Backing Vocals" ]],
+    fn=lambda audio, stem: get_selected_audio(inference(audio), stem),
+    inputs=[audio_input, dropdown], 
+    outputs=audio_output,
     title=title,
     description=description,
+    live=True  # Enable live update for dynamic output
 ).launch()
+
+
+