app.py

import gradio as gr
from audioseal import AudioSeal
import torch
import torchaudio
import traceback

def detect_watermark(audio_file_path):
    try:
        # Load the audio file
        waveform, sample_rate = torchaudio.load(audio_file_path)

        # Ensure waveform has a batch dimension for processing
        if waveform.ndim < 3:
            waveform = waveform.unsqueeze(0)

        # Initialize the AudioSeal detector
        detector = AudioSeal.load_detector("audioseal_detector_16bits")
        
        # Set a conservative threshold for watermark detection
        message_threshold = 0.7  # A higher threshold means more confidence is required to classify as AI-generated
        
        result, confidence = detector.detect_watermark(waveform, message_threshold=message_threshold)

        # Interpret the detection result
        if result:
            detection_result = f"AI-generated with confidence {confidence}"
        else:
            detection_result = "Genuine or the AI watermark is undetectable at the current threshold"

        return f"This audio is likely {detection_result}."
    except Exception as e:
        error_traceback = traceback.format_exc()
        return f"Error occurred: {e}\n\n{error_traceback}"

# Define the Gradio interface
interface = gr.Interface(
    fn=detect_watermark,
    inputs=gr.Audio(label="Upload your audio", type="filepath"),
    outputs="text",
    title="Deep Fake Defender: AI Voice Cloning Detection",
    description="Upload an audio file to check if it's AI-generated or genuine."
)

if __name__ == "__main__":
    interface.launch()