app.py

import os
import gradio as gr
import outetts
import torch
import spaces

# Define available speakers
AVAILABLE_SPEAKERS = [
    "en_male_1", "en_male_2", "en_female_1", "en_female_2",
    "zh_male_1", "zh_male_2", "zh_female_1", "zh_female_2",
    "jp_male_1", "jp_male_2", "jp_female_1", "jp_female_2",
    "kr_male_1", "kr_male_2", "kr_female_1", "kr_female_2"
]

def get_available_speakers():
    return AVAILABLE_SPEAKERS

@spaces.GPU
def generate_tts(text, temperature, repetition_penalty, speaker_selection, reference_audio):
    model_config = outetts.HFModelConfig_v2(
        model_path="OuteAI/OuteTTS-0.3-1B",
        tokenizer_path="OuteAI/OuteTTS-0.3-1B",
        dtype=torch.bfloat16,
        device="cuda"
    )
    interface = outetts.InterfaceHF(model_version="0.3", cfg=model_config)
    
    try:
        if reference_audio:
            speaker = interface.create_speaker(reference_audio)
        elif speaker_selection and speaker_selection != "None":
            speaker = interface.load_default_speaker(speaker_selection)
        else:
            speaker = None
            
        gen_cfg = outetts.GenerationConfig(
            text=text,
            temperature=temperature,
            repetition_penalty=repetition_penalty,
            max_length=4096,
            speaker=speaker,
        )
        output = interface.generate(config=gen_cfg)
        
        if output.audio is None:
            raise ValueError("Audio generation failed. Please try again.")
            
        output_path = "output.wav"
        output.save(output_path)
        return output_path, None
    except Exception as e:
        return None, str(e)

# Custom CSS for 3D effect and modern UI
custom_css = """
.container {
    background: linear-gradient(145deg, #f0f0f0, #ffffff);
    border-radius: 20px;
    box-shadow: 20px 20px 60px #bebebe, -20px -20px 60px #ffffff;
    padding: 2rem;
}
.title {
    font-size: 2.5rem;
    text-align: center;
    background: linear-gradient(45deg, #2196F3, #00BCD4);
    -webkit-background-clip: text;
    -webkit-text-fill-color: transparent;
    margin-bottom: 2rem;
}
.radio-group {
    display: grid;
    grid-template-columns: repeat(auto-fill, minmax(150px, 1fr));
    gap: 1rem;
    margin: 1rem 0;
}
.control-panel {
    background: rgba(255, 255, 255, 0.9);
    border-radius: 15px;
    padding: 1.5rem;
    margin: 1rem 0;
    box-shadow: 0 8px 16px rgba(0,0,0,0.1);
}
.generate-button {
    background: linear-gradient(45deg, #2196F3, #00BCD4);
    color: white;
    border: none;
    padding: 1rem 2rem;
    border-radius: 8px;
    cursor: pointer;
    transition: transform 0.2s;
}
.generate-button:hover {
    transform: translateY(-2px);
}
"""

with gr.Blocks(css=custom_css) as demo:
    with gr.Column(elem_classes="container"):
        gr.Markdown("# Voice Clone Multilingual TTS", elem_classes="title")
        
        with gr.Row():
            with gr.Column(scale=2):
                # Main input section with 3D effect
                with gr.Group(elem_classes="control-panel"):
                    text_input = gr.Textbox(
                        label="Enter Text",
                        placeholder="Type your text here...",
                        lines=3
                    )
                    
                    speaker_radio = gr.Radio(
                        choices=get_available_speakers(),
                        value="en_male_1",
                        label="Choose Voice",
                        elem_classes="radio-group"
                    )
                    
                    with gr.Row():
                        temperature = gr.Slider(
                            minimum=0.1,
                            maximum=1.0,
                            value=0.1,
                            label="Expression Level"
                        )
                        repetition_penalty = gr.Slider(
                            minimum=0.5,
                            maximum=2.0,
                            value=1.1,
                            label="Clarity"
                        )
                    
                    reference_audio = gr.Audio(
                        label="Upload Voice Reference",
                        type="filepath"
                    )
                    
                    submit_button = gr.Button(
                        "Generate Speech",
                        variant="primary",
                        elem_classes="generate-button"
                    )
                    
            with gr.Column(scale=1):
                # Output section with 3D effect
                with gr.Group(elem_classes="control-panel"):
                    audio_output = gr.Audio(
                        label="Generated Audio",
                        type="filepath"
                    )
                    error_box = gr.Textbox(
                        label="Status",
                        visible=False
                    )
        
    submit_button.click(
        fn=generate_tts,
        inputs=[
            text_input,
            temperature,
            repetition_penalty,
            speaker_radio,
            reference_audio,
        ],
        outputs=[audio_output, error_box]
    ).then(
        fn=lambda x: gr.update(visible=bool(x)),
        inputs=[error_box],
        outputs=[error_box]
    )

demo.launch()