import gradio as gr
import os
os.system('cd monotonic_align && python setup.py build_ext --inplace && cd ..')


import json
import math
import torch
from torch import nn
from torch.nn import functional as F
from torch.utils.data import DataLoader

import commons
import utils
from data_utils import TextAudioLoader, TextAudioCollate, TextAudioSpeakerLoader, TextAudioSpeakerCollate
from models import SynthesizerTrn
from text.symbols import symbols
from text import text_to_sequence, cleaned_text_to_sequence
from text.cleaners import japanese_cleaners
from scipy.io.wavfile import write


def get_text(text, hps):
    text_norm = text_to_sequence(text, hps.data.text_cleaners)
    if hps.data.add_blank:
        text_norm = commons.intersperse(text_norm, 0)
    text_norm = torch.LongTensor(text_norm)
    # print(text_norm.shape)
    return text_norm


hps_ms = utils.get_hparams_from_file("configs/japanese_base.json")
hps = utils.get_hparams_from_file("configs/japanese_base.json")
net_g_ms = SynthesizerTrn(
    len(symbols),
    hps_ms.data.filter_length // 2 + 1,
    hps_ms.train.segment_size // hps.data.hop_length,
    n_speakers=hps_ms.data.n_speakers,
    **hps_ms.model)


def jtts(spkid, text):
    sid = torch.LongTensor([spkid])  # speaker identity
    stn_tst = get_text(text, hps_ms)

    with torch.no_grad():
        x_tst = stn_tst.unsqueeze(0)
        x_tst_lengths = torch.LongTensor([stn_tst.size(0)])
        # print(stn_tst.size())
        audio = net_g_ms.infer(x_tst, x_tst_lengths, sid=sid, noise_scale=.667, noise_scale_w=0.8, length_scale=1)[0][
            0, 0].data.float().numpy()
    return

_ = utils.load_checkpoint("output.pth", net_g_ms, None)


def tts(text):
    if len(text) > 150:
        return "Error: Text is too long", None
    sid = torch.LongTensor([2])  # speaker identity
    stn_tst = get_text(text, hps_ms)

    with torch.no_grad():
        x_tst = stn_tst.unsqueeze(0)
        x_tst_lengths = torch.LongTensor([stn_tst.size(0)])
        # print(stn_tst.size())
        audio = net_g_ms.infer(x_tst, x_tst_lengths, sid=sid, noise_scale=.667, noise_scale_w=0.8, length_scale=1)[0][
            0, 0].data.float().numpy()
    return "Success", (hps.data.sampling_rate, audio)

def clean_text(text):
    return japanese_cleaners(text)

def generate_from_clean(text):
    if len(text) > 300:
        return "Error: Text is too long", None
    sid = torch.LongTensor([2])  # speaker identity

    text_norm = cleaned_text_to_sequence(text)
    if hps.data.add_blank:
        text_norm = commons.intersperse(text_norm, 0)
    stn_tst = torch.LongTensor(text_norm)

    with torch.no_grad():
        x_tst = stn_tst.unsqueeze(0)
        x_tst_lengths = torch.LongTensor([stn_tst.size(0)])
        audio = net_g_ms.infer(x_tst, x_tst_lengths, sid=sid, noise_scale=.667, noise_scale_w=0.8, length_scale=1)[0][
            0, 0].data.float().numpy()
    return "Success", (hps.data.sampling_rate, audio)
app = gr.Blocks()
with app:
    with gr.Tabs():
        with gr.TabItem("Basic"):
            tts_input1 = gr.TextArea(label="Text in Japanese (150 words limitation)", value="こんにちは。")
            # tts_input2 = gr.Dropdown(label="Speaker", choices=hps.speakers, type="index", value=hps.speakers[0])
            tts_submit = gr.Button("Generate", variant="primary")
            tts_output1 = gr.Textbox(label="Message")
            tts_output2 = gr.Audio(label="Output")
            tts_submit.click(tts, [tts_input1], [tts_output1, tts_output2])
        with gr.TabItem("Advanced"):
            tts_input3 = gr.TextArea(label="Text in Japanese", value="こんにちは。")
            tts_s1 = gr.Button("Clean", variant="primary")
            tts_input4 = gr.TextArea(label="Cleaned Text (300 words limitation)", value="ko↑Nniʧiwa.")
            tts_s2 = gr.Button("Generate", variant="primary")
            message = gr.Textbox(label="Message")
            tts_o = gr.Audio(label="Output")
            tts_s1.click(clean_text, [tts_input3], [ tts_input4])
            tts_s2.click(generate_from_clean, [tts_input4], [message, tts_o])

    app.launch()