app.py

import gradio as gr
import torchaudio
from tortoise.api import TextToSpeech
from tortoise.utils.audio import load_voice

# Initialize TextToSpeech with the models directory
tts = TextToSpeech(models_dir="urdu-tts/models/urdu-multispeaker")

# Function to generate audio from text
def generate_audio(text, voice, preset):
    try:
        # Load the voice samples and conditioning latents
        voice_samples, conditioning_latents = load_voice(voice)
        
        # Generate the audio
        generated_audio = tts.tts_with_preset(
            text, 
            voice_samples=voice_samples, 
            conditioning_latents=conditioning_latents,
            preset=preset
        )
        
        # Save the generated audio to a file
        output_file = "output.wav"
        torchaudio.save(output_file, generated_audio.squeeze(0).cpu(), 24000)
        
        # Return the audio file
        return output_file
    except Exception as e:
        return f"Error: {str(e)}"

# Define the Gradio interface
with gr.Blocks() as demo:
    gr.Markdown("# Urdu Text-to-Speech")
    
    with gr.Row():
        text_input = gr.Textbox(label="Enter Text (Urdu)", placeholder="محبت، ادب، شاعری، زبان، دلکش")
        voice_dropdown = gr.Dropdown(label="Select Voice", choices=["daniel", "voice2", "voice3"], value="daniel")
        preset_dropdown = gr.Dropdown(label="Select Preset", choices=["fast", "standard", "high_quality"], value="fast")
    
    generate_button = gr.Button("Generate Audio")
    audio_output = gr.Audio(label="Generated Audio")
    
    generate_button.click(generate_audio, inputs=[text_input, voice_dropdown, preset_dropdown], outputs=audio_output)

# Launch the Gradio interface
demo.launch()