app.py

import gradio as gr
import numpy as np
import librosa
import soundfile as sf
from TTS.api import TTS
import torch
import os
import tempfile

# Initialize TTS model
tts = TTS("tts_models/multilingual/multi-dataset/your_tts", progress_bar=False).to("cuda" if torch.cuda.is_available() else "cpu")

def load_audio(audio_path):
    audio, sr = librosa.load(audio_path, sr=None)
    return audio, sr

def save_audio(audio, sr, path):
    sf.write(path, audio, sr)

def pitch_shift(audio, sr, n_steps):
    return librosa.effects.pitch_shift(audio, sr=sr, n_steps=n_steps)

def change_voice(audio_path, pitch_shift_amount, formant_shift_amount):
    # Load the audio
    audio, sr = load_audio(audio_path)
    
    # Apply pitch shifting
    pitched_audio = pitch_shift(audio, sr, pitch_shift_amount)
    
    # Use TTS model for voice conversion
    with tempfile.NamedTemporaryFile(suffix=".wav", delete=False) as temp_file:
        save_audio(pitched_audio, sr, temp_file.name)
        converted_audio_path = tts.voice_conversion(
            source_wav=temp_file.name,
            target_wav="path/to/female_target_voice.wav",  # You need to provide a female target voice file
            output_wav=None
        )
    
    # Load the converted audio
    converted_audio, _ = load_audio(converted_audio_path)
    
    # Apply formant shifting (simplified approach)
    formant_shifted_audio = librosa.effects.pitch_shift(converted_audio, sr=sr, n_steps=formant_shift_amount)
    
    # Clean up temporary files
    os.unlink(temp_file.name)
    os.unlink(converted_audio_path)
    
    return (sr, formant_shifted_audio)

def process_audio(audio_file, pitch_shift_amount, formant_shift_amount):
    sr, audio = change_voice(audio_file.name, pitch_shift_amount, formant_shift_amount)
    
    output_path = "output_voice.wav"
    save_audio(audio, sr, output_path)
    
    return output_path

# Custom CSS for improved design
custom_css = """
.gradio-container {
    background-color: #f0f4f8;
}
.container {
    max-width: 900px;
    margin: auto;
    padding: 20px;
    border-radius: 10px;
    background-color: white;
    box-shadow: 0 4px 6px rgba(0, 0, 0, 0.1);
}
h1 {
    color: #2c3e50;
    text-align: center;
    font-size: 2.5em;
    margin-bottom: 20px;
}
.description {
    text-align: center;
    color: #34495e;
    margin-bottom: 30px;
}
.input-section, .output-section {
    background-color: #ecf0f1;
    padding: 20px;
    border-radius: 8px;
    margin-bottom: 20px;
}
.input-section h3, .output-section h3 {
    color: #2980b9;
    margin-bottom: 15px;
}
"""

# Gradio Interface with improved design
with gr.Blocks(css=custom_css) as demo:
    gr.HTML(
        """
        <div style="text-align: center; max-width: 800px; margin: 0 auto;">
            <div style="display: inline-flex; align-items: center; gap: 0.8rem; font-size: 1.75rem;">
                <svg xmlns="http://www.w3.org/2000/svg" width="1em" height="1em" fill="currentColor" viewBox="0 0 16 16" style="vertical-align: middle;">
                    <path d="M3.5 6.5A.5.5 0 0 1 4 7v1a4 4 0 0 0 8 0V7a.5.5 0 0 1 1 0v1a5 5 0 0 1-4.5 4.975V15h3a.5.5 0 0 1 0 1h-7a.5.5 0 0 1 0-1h3v-2.025A5 5 0 0 1 3 8V7a.5.5 0 0 1 .5-.5z"/>
                    <path d="M10 8a2 2 0 1 1-4 0V3a2 2 0 1 1 4 0v5zM8 0a3 3 0 0 0-3 3v5a3 3 0 0 0 6 0V3a3 3 0 0 0-3-3z"/>
                </svg>
                <h1 style="font-weight: 900; margin-bottom: 7px;">
                    AI Voice Changer
                </h1>
            </div>
            <p class="description">Transform any voice into a realistic female voice using advanced AI technology</p>
        </div>
        """
    )
    
    with gr.Row():
        with gr.Column(elem_classes="input-section"):
            gr.Markdown("### Input")
            audio_input = gr.Audio(type="filepath", label="Upload Voice")
            pitch_shift = gr.Slider(-12, 12, step=0.5, label="Pitch Shift", value=0)
            formant_shift = gr.Slider(-5, 5, step=0.1, label="Formant Shift", value=0)
            submit_btn = gr.Button("Transform Voice", variant="primary")

        with gr.Column(elem_classes="output-section"):
            gr.Markdown("### Output")
            audio_output = gr.Audio(label="Transformed Voice")

    submit_btn.click(
        fn=process_audio,
        inputs=[audio_input, pitch_shift, formant_shift],
        outputs=audio_output,
    )

    gr.Markdown(
        """
        ### How to use:
        1. Upload an audio file containing the voice you want to transform.
        2. Adjust the Pitch Shift and Formant Shift sliders to fine-tune the voice (optional).
        3. Click the "Transform Voice" button to process the audio.
        4. Listen to the transformed voice in the output section.
        5. Download the transformed audio file if desired.

        Note: This application uses AI to transform voices. The quality of the output may vary depending on the input audio quality and the chosen settings.
        """
    )

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