import gradio as gr
from transformers import VitsModel, AutoTokenizer
import torch
import scipy.io.wavfile

# Initialize TTS models
tts = TTS(model_name="tts_models/multilingual/multi-dataset/your_tts", progress_bar=False, gpu=False)
zh_tts = TTS(model_name="tts_models/zh-CN/baker/tacotron2-DDC-GST", progress_bar=False, gpu=False)
de_tts = TTS(model_name="tts_models/de/thorsten/vits", gpu=False)
es_tts = TTS(model_name="tts_models/es/mai/tacotron2-DDC", progress_bar=False, gpu=False)
tam_tts_model = VitsModel.from_pretrained("facebook/mms-tts-tam")
tam_tokenizer = AutoTokenizer.from_pretrained("facebook/mms-tts-tam")

def text_to_speech(text: str, speaker_wav, speaker_wav_file, language: str):
    if speaker_wav_file and not speaker_wav:
        speaker_wav = speaker_wav_file
    file_path = "output.wav"
    if language == "zh-CN":
        zh_tts.tts_to_file(text, file_path=file_path)
    elif language == "de":
        de_tts.tts_to_file(text, file_path=file_path)
    elif language == "es":
        es_tts.tts_to_file(text, file_path=file_path)
    elif language == "tam":
        inputs = tam_tokenizer(text, return_tensors="pt")
        with torch.no_grad():
            output = tam_tts_model(**inputs).waveform
        scipy.io.wavfile.write(file_path, rate=tam_tts_model.config.sampling_rate, data=output.numpy())
    else:
        if speaker_wav is not None:
            tts.tts_to_file(text, speaker_wav=speaker_wav, language=language, file_path=file_path)
        else:
            tts.tts_to_file(text, speaker=tts.speakers[0], language=language, file_path=file_path)
    return file_path

title = "Voice-Cloning-Demo"

def toggle(choice):
    if choice == "mic":
        return gr.update(visible=True, value=None), gr.update(visible=False, value=None)
    else:
        return gr.update(visible=False, value=None), gr.update(visible=True, value=None)

def handle_language_change(choice):
    if choice in ["zh-CN", "de", "es", "tam"]:
        return gr.update(visible=False), gr.update(visible=False), gr.update(visible(False))
    else:
        return gr.update(visible=True), gr.update(visible(True)), gr.update(visible(True))

warming_text = """Please note that Chinese, German, Spanish, and Tamil are currently not supported for voice cloning."""

with gr.Blocks() as demo:
    with gr.Row():
        with gr.Column():
            text_input = gr.Textbox(label="Input the text", value="", max_lines=3)
            lan_input = gr.Radio(label="Language", choices=["en", "fr-fr", "pt-br", "zh-CN", "de", "es", "tam"], value="en")
            gr.Markdown(warming_text)
            radio = gr.Radio(["mic", "file"], value="mic",
                             label="How would you like to upload your audio?")
            audio_input_mic = gr.Audio(label="Voice to clone", source="microphone", type="filepath", visible=True)
            audio_input_file = gr.Audio(label="Voice to clone", type="filepath", visible=False)

            with gr.Row():
                with gr.Column():
                    btn_clear = gr.Button("Clear")
                with gr.Column():
                    btn = gr.Button("Submit", variant="primary")
        with gr.Column():
            audio_output = gr.Audio(label="Output")

    btn.click(text_to_speech, inputs=[text_input, audio_input_mic,
              audio_input_file, lan_input], outputs=audio_output)
    radio.change(toggle, radio, [audio_input_mic, audio_input_file])
    lan_input.change(handle_language_change, lan_input, [radio, audio_input_mic, audio_input_file])

demo.launch(enable_queue=True)