app.py

import gradio as gr
from TTS.api import TTS
import numpy as np
from scipy.io import wavfile
import tempfile
import os

# Load the YourTTS model once at startup
tts = TTS(model_name="tts_models/multilingual/multi-dataset/your_tts", progress_bar=False)
sample_rate = tts.synthesizer.output_sample_rate

def generate_speech(reference_audio, text):
    """
    Generate speech audio mimicking the voice from the reference audio.
    
    Parameters:
    reference_audio (str): Filepath to the uploaded voice sample.
    text (str): Text to convert to speech.
    
    Returns:
    str: Path to the generated audio file
    """
    # Generate speech using the reference audio and text
    wav = tts.tts(text=text, speaker_wav=reference_audio, language="en")
    # Convert list to numpy array
    wav_np = np.array(wav, dtype=np.float32)
    
    # Create a temporary file to save the audio
    temp_file = tempfile.NamedTemporaryFile(suffix=".wav", delete=False)
    temp_file_path = temp_file.name
    # Save the audio to the temporary file
    wavfile.write(temp_file_path, sample_rate, wav_np)
    temp_file.close()
    
    return temp_file_path

# Build the Gradio interface
with gr.Blocks(title="Voice Cloning TTS") as app:
    gr.Markdown("## Voice Cloning Text-to-Speech")
    gr.Markdown("Upload a short voice sample in English, then enter text to hear it in your voice!")
    
    with gr.Row():
        audio_input = gr.Audio(type="filepath", label="Upload Your Voice Sample (English)")
        text_input = gr.Textbox(label="Enter Text to Convert to Speech", placeholder="e.g., I love chocolate")
    
    generate_btn = gr.Button("Generate Speech")
    audio_output = gr.Audio(label="Generated Speech", interactive=False)
    
    # Connect the button to the generation function
    generate_btn.click(
        fn=generate_speech,
        inputs=[audio_input, text_input],
        outputs=audio_output
    )

# Launch the application
app.launch()