app.py

import gradio as gr
import tempfile
import openai
import requests
import os

def tts(input_text: str, model: str, voice: str, api_key: str) -> str:
    """
    Convert input text to speech using OpenAI's Text-to-Speech API.

    Parameters:
        input_text (str): The text to be converted to speech.
        model (str): The model to use for synthesis (e.g., 'tts-1', 'tts-1-hd').
        voice (str): The voice profile to use (e.g., 'alloy', 'echo', 'fable', etc.).
        api_key (str): OpenAI API key.

    Returns:
        str: File path to the generated audio file.

    Raises:
        gr.Error: If input parameters are invalid or API call fails.
    """
    if not api_key.strip():
        raise gr.Error(
            "API key is required. Get an API key at: https://platform.openai.com/account/api-keys"
        )

    if not input_text.strip():
        raise gr.Error("Input text cannot be empty.")

    openai.api_key = api_key

    try:
        response = openai.Audio.create(text=input_text, voice=voice, model=model)
    except openai.OpenAIError as e:
        # Catch-all for OpenAI exceptions
        raise gr.Error(f"An OpenAI error occurred: {e}")
    except Exception as e:
        # Catch any other exceptions
        raise gr.Error(f"An unexpected error occurred: {e}")

    if not hasattr(response, "audio"):
        raise gr.Error(
            "Invalid response from OpenAI API. The response does not contain audio content."
        )

    with tempfile.NamedTemporaryFile(suffix=".mp3", delete=False) as temp_file:
        temp_file.write(response.audio)
        temp_file_path = temp_file.name

    return temp_file_path

def main():
    """
    Main function to create and launch the Gradio interface.
    """
    MODEL_OPTIONS = ["tts-1", "tts-1-hd"]
    VOICE_OPTIONS = ["alloy", "echo", "fable", "onyx", "nova", "shimmer"]

    # Predefine voice previews URLs
    VOICE_PREVIEW_URLS = {
        voice: f"https://cdn.openai.com/API/docs/audio/{voice}.wav"
        for voice in VOICE_OPTIONS
    }

    # Download audio previews to disk before initiating the interface
    PREVIEW_DIR = "voice_previews"
    os.makedirs(PREVIEW_DIR, exist_ok=True)

    VOICE_PREVIEW_FILES = {}
    for voice, url in VOICE_PREVIEW_URLS.items():
        local_file_path = os.path.join(PREVIEW_DIR, f"{voice}.wav")
        if not os.path.exists(local_file_path):
            try:
                response = requests.get(url)
                response.raise_for_status()
                with open(local_file_path, "wb") as f:
                    f.write(response.content)
            except requests.exceptions.RequestException as e:
                print(f"Failed to download {voice} preview: {e}")
        VOICE_PREVIEW_FILES[voice] = local_file_path

    # Set static paths for Gradio to serve
    # This needs to be done before creating the Gradio app
    gr.set_static_paths([PREVIEW_DIR])

    with gr.Blocks(title="OpenAI - Text to Speech") as demo:
        with gr.Row():
            with gr.Column(scale=1):
                gr.Markdown("### Voice Previews")

                # A function to update the preview_audio component
                def play_voice_sample(voice):
                    return gr.update(value=VOICE_PREVIEW_FILES[voice])

                # Create buttons for each voice inside a grid
                for voice in VOICE_OPTIONS:
                    # Create a button for each voice
                    voice_button = gr.Button(
                        value=f"{voice.capitalize()}",
                        variant="secondary",
                        size="sm",
                    )

                    # Attach the click handler
                    voice_button.click(
                        fn=lambda v=voice: play_voice_sample(v),
                        outputs=preview_audio,
                    )

                # Create an audio component to play the samples
                preview_audio = gr.Audio(
                    interactive=False,
                    label="Preview Audio",
                    value=None,
                    visible=True,
                    autoplay=True,
                )

            with gr.Column(scale=1):
                api_key_input = gr.Textbox(
                    label="OpenAI API Key",
                    info="https://platform.openai.com/account/api-keys",
                    type="password",
                    placeholder="Enter your OpenAI API Key",
                )
                model_dropdown = gr.Dropdown(
                    choices=MODEL_OPTIONS,
                    label="Model",
                    value="tts-1",
                )
                voice_dropdown = gr.Dropdown(
                    choices=VOICE_OPTIONS,
                    label="Voice Options",
                    value="echo",
                )

            with gr.Column(scale=2):
                input_textbox = gr.Textbox(
                    label="Input Text",
                    lines=10,
                    placeholder="Type your text here...",
                )
                submit_button = gr.Button(
                    "Convert Text to Speech",
                    variant="primary",
                )
            with gr.Column(scale=1):
                output_audio = gr.Audio(label="Output Audio")

        # Define the event handler for the submit button with error handling
        def on_submit(input_text, model, voice, api_key):
            audio_file = tts(input_text, model, voice, api_key)
            return audio_file

        # Trigger the conversion when the submit button is clicked
        submit_button.click(
            fn=on_submit,
            inputs=[input_textbox, model_dropdown, voice_dropdown, api_key_input],
            outputs=output_audio,
        )

    # Launch the Gradio app with error display enabled
    demo.launch(show_error=True)

if __name__ == "__main__":
    main()