app.py

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 as symbols_default 

from scipy.io.wavfile import write
from text import cleaners

model_configs = {
    "Graphemes": {
        "path": "french_model_vits/G_700000.pth",
        "symbols": symbols_default
    }
}

# Global variables 
net_g = None
symbols = []
_symbol_to_id = {}
_id_to_symbol = {}

def text_to_sequence(text, cleaner_names):
    sequence = []
    clean_text = _clean_text(text, cleaner_names)
    for symbol in clean_text:
        symbol_id = _symbol_to_id[symbol]
        sequence += [symbol_id]
    return sequence

def _clean_text(text, cleaner_names):
    for name in cleaner_names:
        cleaner = getattr(cleaners, name)
        if not cleaner:
            raise Exception('Unknown cleaner: %s' % name)
        text = cleaner(text)
    return text

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)
    return text_norm

def load_model_and_symbols(tab_name):
    global net_g, symbols, _symbol_to_id, _id_to_symbol
    model_config = model_configs[tab_name]
    symbols = model_config["symbols"]
    _symbol_to_id = {s: i for i, s in enumerate(symbols)}
    _id_to_symbol = {i: s for i, s in enumerate(symbols)}

    net_g = SynthesizerTrn(
        len(symbols),
        hps.data.filter_length // 2 + 1,
        hps.train.segment_size // hps.data.hop_length,
        n_speakers=hps.data.n_speakers,
        **hps.model)
    _ = net_g.eval()
    _ = utils.load_checkpoint(model_config["path"], net_g, None)

def tts(text, speaker_id, tab_name):
    load_model_and_symbols(tab_name)
    sid = torch.LongTensor([speaker_id])  # speaker identity
    stn_tst = get_text(text, hps)

    with torch.no_grad():
        x_tst = stn_tst.unsqueeze(0)
        x_tst_lengths = torch.LongTensor([stn_tst.size(0)])
        audio = net_g.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 create_tab(tab_name):
    with gr.TabItem(tab_name):
        gr.Markdown(f"### {tab_name} TTS Model")
        tts_input1 = gr.TextArea(label="Text in french", value="")
        tts_input2 = gr.Dropdown(label="Speaker", choices=["Male", "Female"], type="index", value="Male")
        tts_submit = gr.Button("Generate", variant="primary")
        tts_output1 = gr.Textbox(label="Message")
        tts_output2 = gr.Audio(label="Output")
        tts_submit.click(lambda text, speaker_id: tts(text, speaker_id, tab_name), [tts_input1, tts_input2], [tts_output1, tts_output2])

hps = utils.get_hparams_from_file("configs/vctk_base.json")

app = gr.Blocks()
with app:
    gr.Markdown(
        """
        # VITS Implementation for French
        
        Based on VITS (https://github.com/jaywalnut310/vits).
        
        ## How to use:
        Write the text on the box below. For faster inference, it is recommended to use short sentences.
        
        ## Hint: Some sample texts are available at the bottom of the web site.
        """
    )
    with gr.Tabs():
        create_tab("French TTS")

    gr.Markdown(
        """
        ## Examples
        | Input Text | Speaker |
        |------------|---------|
        | On ne voit bien qu'avec le cœur, l'essentiel est invisible pour les yeux. | Female | 
        | Voilà plusieurs fois, Monsieur, que je vous rencontre sur mon chemin. C’est autant de fois de trop, et j’en ai assez de perdre mon temps à déjouer les pièges que vous me tendez. | Male |
        | Je pense, donc je suis.  | Female | 
        | La vie est un sommeil, l'amour en est le rêve, et vous aurez vécu si vous avez aimé. | Male | 
        """
    )

app.launch()