Create app.py

app.py

@@ -0,0 +1,50 @@
+import gradio as gr
+import soundfile as sf
+from huggingface_hub import hf_hub_download
+from audiosr.inference import super_resolution
+import numpy as np
+import os
+
+# Download the OpenVINO models
+hf_hub_download(repo_id="Intel/versatile_audio_super_resolution_openvino", filename="versatile_audio_sr_base_openvino_models.zip", local_dir=".")
+import zipfile
+with zipfile.ZipFile("versatile_audio_sr_base_openvino_models.zip", 'r') as zip_ref:
+    zip_ref.extractall(".")
+
+# Define the model paths
+model_paths = {
+    'audio_sr_decoder': 'audio_sr_decoder.xml',
+    'audio_sr_encoder': 'audio_sr_encoder.xml',
+    'vae_feature_extract': 'vae_feature_extract.xml',
+    'vocoder': 'vocoder.xml'
+}
+
+def audio_super_resolution(audio_file):
+    """
+    Performs audio super-resolution on the input audio file.
+    """
+    waveform, sr = sf.read(audio_file)
+    if len(waveform.shape) > 1:
+        waveform = np.mean(waveform, axis=1) # apects mono audio
+    sf.write("input.wav", waveform, sr)
+
+    # Perform super-resolution
+    output = super_resolution(
+        "input.wav",
+        "output.wav",
+        model_paths=model_paths
+    )
+    return "output.wav"
+
+# Create the Gradio interface
+iface = gr.Interface(
+    fn=audio_super_resolution,
+    inputs=gr.Audio(type="filepath", label="Input Audio"),
+    outputs=gr.Audio(type="filepath", label="Output Audio"),
+    title="Versatile Audio Super Resolution (OpenVINO)",
+    description="Upload an audio file to perform super-resolution.",
+    examples=[["example.wav"]]
+)
+
+if __name__ == "__main__":
+    iface.launch()