app.py

import os
import gradio as gr
import outetts
from outetts.version.v2.interface import _DEFAULT_SPEAKERS
import torch
import spaces

def get_available_speakers():
    speakers = list(_DEFAULT_SPEAKERS.keys())
    return 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:
        # Validate inputs for custom speaker
        if reference_audio:
            speaker = interface.create_speaker(reference_audio)
        # Use selected default speaker
        elif speaker_selection and speaker_selection != "None":
            speaker = interface.load_default_speaker(speaker_selection)
        # No speaker - random characteristics
        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)

        # Verify output
        if output.audio is None:
            raise ValueError("Model failed to generate audio. This may be due to input length constraints or early EOS token.")

        # Save and return output
        output_path = "output.wav"
        output.save(output_path)
        return output_path, None
    except Exception as e:
        return None, str(e)

# Custom CSS for 3D styling
custom_css = """
.container {
    background: linear-gradient(145deg, #f3f4f6, #ffffff);
    border-radius: 20px;
    box-shadow: 10px 10px 20px #d1d1d1, -10px -10px 20px #ffffff;
    padding: 2rem;
    margin: 1rem;
    transition: all 0.3s ease;
    min-height: 800px;
    width: 100%;
    max-width: 1400px;
    margin: 0 auto;
}

.title {
    font-size: 2.5rem;
    font-weight: bold;
    color: #1a1a1a;
    text-align: center;
    margin-bottom: 2rem;
    text-shadow: 2px 2px 4px rgba(0, 0, 0, 0.1);
}

.input-group {
    background: #ffffff;
    border-radius: 15px;
    padding: 1.5rem;
    margin: 1rem 0;
    box-shadow: inset 5px 5px 10px #e0e0e0, inset -5px -5px 10px #ffffff;
}

.button-3d {
    background: linear-gradient(145deg, #3b82f6, #2563eb);
    color: white;
    border: none;
    padding: 0.8rem 1.5rem;
    border-radius: 10px;
    font-weight: bold;
    cursor: pointer;
    transition: all 0.3s ease;
    box-shadow: 5px 5px 10px #d1d1d1, -5px -5px 10px #ffffff;
}

.button-3d:hover {
    transform: translateY(-2px);
    box-shadow: 7px 7px 15px #d1d1d1, -7px -7px 15px #ffffff;
}

.slider-3d {
    height: 12px;
    border-radius: 6px;
    background: linear-gradient(145deg, #e6e7eb, #ffffff);
    box-shadow: inset 3px 3px 6px #d1d1d1, inset -3px -3px 6px #ffffff;
}

.error-box {
    background: #fee2e2;
    border-left: 4px solid #ef4444;
    padding: 1rem;
    border-radius: 8px;
    margin: 1rem 0;
}

.right-column {
    display: flex;
    flex-direction: column;
    gap: 1rem;
}

.options-panel {
    margin-top: 2rem;
    background: linear-gradient(145deg, #f3f4f6, #ffffff);
    border-radius: 15px;
    padding: 2.5rem;
    box-shadow: 5px 5px 10px #d1d1d1, -5px -5px 10px #ffffff;
    display: flex;
    flex-direction: column;
    gap: 2rem;
}

.input-box {
    padding: 1.5rem;
    background: #ffffff;
    border-radius: 12px;
    margin-bottom: 1.5rem;
    box-shadow: inset 3px 3px 7px #e0e0e0, inset -3px -3px 7px #ffffff;
}

.slider-container {
    margin: 2rem 0;
}

.textbox-container {
    min-height: 150px;
}
"""

# Create the Gradio interface with 3D styling
with gr.Blocks(css=custom_css) as demo:
    gr.Markdown('<div class="title">Voice Clone Multilingual TTS</div>')
    
    error_box = gr.Textbox(label="Error Messages", visible=False, elem_classes="error-box")
    
    with gr.Row(elem_classes="container"):
        # Left column for text input
        with gr.Column(scale=1):
            with gr.Group(elem_classes="textbox-container"):
                text_input = gr.Textbox(
                    label="Text to Synthesize",
                    placeholder="Enter text here...",
                    elem_classes="input-group",
                    lines=8
                )
            
            submit_button = gr.Button(
                "Generate Speech",
                elem_classes="button-3d"
            )

        # Right column for output and options
        with gr.Column(scale=1, elem_classes="right-column"):
            # Audio output at the top
            audio_output = gr.Audio(
                label="Generated Audio",
                type="filepath",
                elem_classes="input-group"
            )
            
            # Options panel below the output
            with gr.Group(elem_classes="options-panel"):
                speaker_dropdown = gr.Dropdown(
                    choices=get_available_speakers(),
                    value="en_male_1",
                    label="Speaker Selection",
                    elem_classes="input-group"
                )
                
                with gr.Group(elem_classes="slider-container"):
                    temperature = gr.Slider(
                        0.1, 1.0,
                        value=0.1,
                        label="Temperature (lower = more stable tone, higher = more expressive)",
                        elem_classes="slider-3d"
                    )
                
                with gr.Group(elem_classes="slider-container"):
                    repetition_penalty = gr.Slider(
                        0.5, 2.0,
                        value=1.1,
                        label="Repetition Penalty",
                        elem_classes="slider-3d"
                    )
                
                reference_audio = gr.Audio(
                    label="Reference Audio (for voice cloning)",
                    type="filepath",
                    elem_classes="input-group"
                )
                
                gr.Markdown("""
                ### Voice Cloning Guidelines:
                - Use around 7-10 seconds of clear, noise-free audio
                - For transcription interface will use Whisper turbo to transcribe the audio file
                - Longer audio clips will reduce maximum output length
                - Custom speaker overrides speaker selection
                """, elem_classes="input-group")
    
    submit_button.click(
        fn=generate_tts,
        inputs=[
            text_input,
            temperature,
            repetition_penalty,
            speaker_dropdown,
            reference_audio,
        ],
        outputs=[audio_output, error_box]
    ).then(
        fn=lambda x: gr.update(visible=bool(x)),
        inputs=[error_box],
        outputs=[error_box]
    )

demo.launch()