import random 
import gradio as gr 
import numpy as np 
from elevenlabs import voices, generate, set_api_key, UnauthenticatedRateLimitError

def pad_buffer(audio):
    # Pad buffer to multiple of 2 bytes
    buffer_size = len(audio)
    element_size = np.dtype(np.int16).itemsize
    if buffer_size % element_size != 0:
        audio = audio + b'\0' * (element_size - (buffer_size % element_size))
    return audio 

def generate_voice(text, voice_name):
    try:
        audio = generate(
            text[:250], # Limit to 250 characters
            voice=voice_name, 
            model="eleven_multilingual_v2"
        )
        return (44100, np.frombuffer(pad_buffer(audio), dtype=np.int16))
    except UnauthenticatedRateLimitError as e:
        raise gr.Error("Thanks for trying out ElevenLabs TTS! You've reached the free tier limit. Please provide an API key to continue.") 
    except Exception as e:
        raise gr.Error(e)
    

description = """
Here's a demonstration of the world's most advanced TTS systems, created by ElevenLabs. 🎉 I wanted to experiment with this amazing technology for the Italian language 🇮🇹, and I'm excited to share its capabilities with you! Eleven Multilingual V2 is a single foundational model supporting an impressive 28 languages, including English 🇬🇧, Chinese 🇨🇳, Spanish 🇪🇸, Hindi 🇮🇳, Portuguese 🇵🇹, French 🇫🇷, German 🇩🇪, Japanese 🇯🇵, Arabic 🇸🇦, Korean 🇰🇷, Indonesian 🇮🇩, Italian 🇮🇹, Dutch 🇳🇱, Turkish 🇹🇷, Polish 🇵🇱, Swedish 🇸🇪, Filipino 🇵🇭, Malay 🇲🇾, Romanian 🇷🇴, Ukrainian 🇺🇦, Greek 🇬🇷, Czech 🇨🇿, Danish 🇩🇰, Finnish 🇫🇮, Bulgarian 🇧🇬, Croatian 🇭🇷, Slovak 🇸🇰, and Tamil 🇱🇰. 🌍 Sign up on ElevenLabs to get fast access to long-form generation, voice cloning, API keys, and more! 🚀
"""

with gr.Blocks() as block:
    gr.Markdown('[ ![ElevenLabs](https://user-images.githubusercontent.com/12028621/262629275-4f85c9cf-85b6-435e-ab50-5b8c7c4e9dd2.png) ](https://elevenlabs.io)')
    gr.Markdown(description)
    
    input_text = gr.Textbox(
        label="Input Text (250 characters max)", 
        lines=2, 
        value="Ciao, mi chiamo Sab, sono un ragazzo italiano appassionato di AI e nuove tecnologie!",
        elem_id="input_text"
    )

    all_voices = voices() 
    input_voice = gr.Dropdown(
        [ voice.name for voice in all_voices ], 
        value="Callum",
        label="Voice", 
        elem_id="input_voice"
    )

    run_button = gr.Button(
        value="Generate Voice", 
    )

    out_audio = gr.Audio(
        label="Generated Voice",
        type="numpy", 
        elem_id="out_audio",
        format="mp3"
    )
        
    inputs = [input_text, input_voice]
    outputs = [out_audio]
    
    run_button.click(
        fn=generate_voice, 
        inputs=inputs, 
        outputs=outputs, 
        queue=True
    )

block.launch(debug=False, show_error=True, share=True)