import gradio as gr
from transformers import pipeline

# Load the pre-trained voice conversion model
# You can replace with any RVC / voice conversion model from Hugging Face
vc_pipeline = pipeline("audio-to-audio", model="wok000/RVC-Hindi-Voice")

def convert_voice(input_audio):
    # input_audio is a tuple: (sample_rate, numpy_array)
    sr, data = input_audio
    # Save the input to a temporary file for processing
    input_path = "input.wav"
    import soundfile as sf
    sf.write(input_path, data, sr)

    # Run voice conversion
    result = vc_pipeline(input_path)
    
    # The pipeline returns a dictionary with 'audio'
    output_audio_path = "output.wav"
    with open(output_audio_path, "wb") as f:
        f.write(result["audio"])
    
    return output_audio_path


# Gradio Interface
demo = gr.Interface(
    fn=convert_voice,
    inputs=gr.Audio(sources=["microphone", "upload"], type="numpy", label="Upload or Record Audio"),
    outputs=gr.Audio(type="file", label="Converted Voice"),
    title="Indian Accent Voice Conversion",
    description="Upload a voice sample and convert it to a Hindi-styled accent using a pre-trained RVC model."
)

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