Update app.py

app.py

@@ -7,7 +7,7 @@ import tempfile
 from demucs.pretrained import get_model
 from demucs.apply import apply_model
 
-# Load the Demucs model
+# Load Demucs model (htdemucs is good for vocals removal)
 model = get_model("htdemucs")
 model.cpu()
 model.eval()
@@ -18,14 +18,14 @@ def remove_vocals(audio_file):
     if y.ndim == 1:
         y = np.expand_dims(y, axis=0)
 
-    audio_tensor = torch.tensor(y).unsqueeze(0)  # shape (1, channels, samples)
+    audio_tensor = torch.tensor(y).unsqueeze(0)  # shape: (1, channels, samples)
 
     with torch.no_grad():
-        sources = apply_model(model, audio_tensor)  # Demucs returns 5 sources: [drums, bass, other, vocals, remix]
+        # Demucs separates: [drums, bass, other, vocals, remix]
+        sources = apply_model(model, audio_tensor)
 
-    # Sources indices for htdemucs: 0-drums, 1-bass, 2-other, 3-vocals, 4-mix
-    # We want music + expression (everything except clean vocals)
-    music_expression = sources[0][[0,1,2]].sum(dim=0)  # drums+bass+other
+    # Keep music + expression sounds (everything except clean vocals)
+    music_expression = sources[0][[0,1,2]].sum(dim=0)  # drums + bass + other
 
     # Save output
     out_path = tempfile.NamedTemporaryFile(delete=False, suffix=".wav").name
@@ -37,8 +37,8 @@ def remove_vocals(audio_file):
 interface = gr.Interface(
     fn=remove_vocals,
     inputs=gr.Audio(type="filepath", label="Upload Audio"),
-    outputs=gr.Audio(label="Background Music + Expressions"),
-    title="Vocal Remover (Keep Music + Expressions)"
+    outputs=gr.Audio(label="Music + Expression Sounds"),
+    title="Vocal Remover: Keep Music + Expression Sounds"
 )
 
 interface.launch()