import gradio as gr
import os
import tempfile
from openai import OpenAI


def generate_tts(text, model, voice, api_key,response_format, speed):
    if api_key == '':
        raise gr.Error('Please enter your OpenAI API key to run this generator')
    else:
        try:
            client = OpenAI(api_key=api_key)

            response = client.audio.speech.create(
                model=model, # tts-1, tts-1-hd
                voice=voice, # alloy, echo, fable, onyx, nova, shimmer
                response_format=response_format, # mp3, opus, aac, flac
                speed=speed, # 0.25x - 4x
                input=text,
            )

        except Exception as error:
            # Handle any exception that occurs
            raise gr.Error("An error occurred while generating speech output. Please check your API key and try again.")
            print(str(error))

    # Create a temp file to save the audio from OpenAI api
    output_suffix = "." + response_format
    with tempfile.NamedTemporaryFile(suffix=output_suffix, delete=False) as temp_file:
        temp_file.write(response.content)

    # Get the file path of the temp file
    temp_file_path = temp_file.name

    return temp_file_path

css = """
.header-text p {line-height: 80px !important; text-align: left; font-size: 26px;}
.header-logo {text-align: left}
"""

with gr.Blocks(css=css) as tts_demo:
    with gr.Row():
        with gr.Column(scale=1, min_width=80):
            gr.Image("tt-logo.jpg", width=80, height=80, show_download_button=False, show_share_button=False, interactive=False, show_label=False, elem_id="thinktecture-logo", elem_classes="header-logo", container=False)
        with gr.Column(scale=11):
            gr.Markdown("OpenAI Text-To-Speech Generator (TTS-1 Model)", elem_classes="header-text")
        
    with gr.Row(variant='panel'):
      api_key = gr.Textbox(type='password', label='OpenAI API Key', placeholder='Your OpenAI API key')
      model = gr.Dropdown(choices=['tts-1','tts-1-hd'], label='Model', value='tts-1')
      voice = gr.Dropdown(choices=['alloy', 'echo', 'fable', 'onyx', 'nova', 'shimmer'], label='Voice', value='fable')
      response_format = gr.Dropdown(choices=['mp3', 'opus', 'aac', 'flac'], label='Output format', value='mp3')
      speed = gr.Slider(0.25, 4, value=1, step=.25, label="Speed", info="Speach speed (0.25x - 4x)")

    with gr.Row(variant='panel'):
            text = gr.Textbox(label="Input text", placeholder="Your input for Text-To-Speech - press 'Say it' or hit 'Enter' key when ready.")
    with gr.Row(variant='panel'):
        with gr.Column(scale=1):
            gr.Markdown(" ")
        with gr.Column(scale=1):
            btn = gr.Button("Say it")
        with gr.Column(scale=1):
            gr.Markdown(" ")
        
    output_audio = gr.Audio(label="Text-To-Speech Output", waveform_options={"waveform_progress_color" : "#ff584f"})
    
    text.submit(fn=generate_tts, inputs=[text, model, voice, api_key, response_format, speed], outputs=output_audio, api_name="tts_enter_key", concurrency_limit=None)
    btn.click(fn=generate_tts, inputs=[text, model, voice, api_key, response_format, speed], outputs=output_audio, api_name="tts_button", concurrency_limit=None)

tts_demo.launch()