Create app.py

app.py

@@ -0,0 +1,57 @@
+import gradio as gr
+import torch
+import tempfile
+import soundfile as sf
+import librosa
+import numpy as np
+
+from demucs.pretrained import get_model
+from demucs.apply import apply_model
+
+# Load Demucs model
+model = get_model("htdemucs")
+model.cpu()
+model.eval()
+
+
+def remove_vocals(audio_file):
+
+    # load audio
+    audio, sr = librosa.load(audio_file, sr=44100, mono=False)
+
+    if audio.ndim == 1:
+        audio = np.expand_dims(audio, axis=0)
+
+    audio_tensor = torch.tensor(audio).unsqueeze(0)
+
+    with torch.no_grad():
+        sources = apply_model(model, audio_tensor)
+
+    # sources order:
+    # 0 drums
+    # 1 bass
+    # 2 other
+    # 3 vocals
+
+    drums = sources[0][0].cpu().numpy()
+    bass = sources[0][1].cpu().numpy()
+    other = sources[0][2].cpu().numpy()
+
+    # mix everything except vocals
+    background = drums + bass + other
+
+    output_file = tempfile.NamedTemporaryFile(delete=False, suffix=".wav").name
+
+    sf.write(output_file, background.T, sr)
+
+    return output_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)"
+)
+
+interface.launch()