Create app.py

app.py

@@ -0,0 +1,40 @@
+import gradio as gr
+import torch
+import torchaudio
+from transformers import AutoProcessor, AutoModelForAudioClassification
+
+# Load the Hugging Face processor and model for audio deepfake detection.
+processor = AutoProcessor.from_pretrained("MelodyMachine/Deepfake-audio-detection-V2")
+model = AutoModelForAudioClassification.from_pretrained("MelodyMachine/Deepfake-audio-detection-V2")
+
+def detect_deepfake_audio(audio_path: str) -> str:
+    # Load audio (supports WAV, MP3, FLAC, etc.)
+    waveform, sample_rate = torchaudio.load(audio_path)
+
+    # Convert to mono if necessary
+    if waveform.shape[0] > 1:
+        waveform = torch.mean(waveform, dim=0, keepdim=True)
+
+    # Preprocess for model
+    inputs = processor(waveform, sampling_rate=sample_rate, return_tensors="pt")
+    with torch.no_grad():
+        outputs = model(**inputs)
+
+    # Get prediction
+    probs = torch.softmax(outputs.logits, dim=-1)[0]
+    idx = torch.argmax(probs).item()
+    label = model.config.id2label[idx]
+    confidence = probs[idx].item()
+
+    return f"The audio is classified as **{label}** with confidence **{confidence:.2f}**"
+
+# Build Gradio interface
+with gr.Blocks() as demo:
+    gr.Markdown("# Audio Deepfake Detection App")
+    gr.Markdown("### Upload or record an audio clip to detect deepfake content.")
+    audio_in = gr.Audio(source="upload", type="filepath", label="Upload Audio")
+    txt_out = gr.Textbox(label="Result")
+    gr.Button("Detect").click(fn=detect_deepfake_audio, inputs=audio_in, outputs=txt_out)
+
+if __name__ == "__main__":
+    demo.launch()