Skip to content

Instantly share code, notes, and snippets.

What would you like to do?
import os
import sys
import io
import torch
from collections import OrderedDict
from TTS.models.tacotron import Tacotron
from TTS.layers import *
from import *
from import AudioProcessor
from TTS.utils.generic_utils import load_config
from TTS.utils.text import text_to_sequence
from TTS.utils.synthesis import synthesis
from utils.text.symbols import symbols, phonemes
from TTS.utils.visual import visualize
# Set constants
MODEL_PATH = './tts_model/best_model.pth.tar'
CONFIG_PATH = './tts_model/config.json'
OUT_FILE = 'tts_out.wav'
CONFIG = load_config(CONFIG_PATH)
use_cuda = False
def tts(model, text, CONFIG, use_cuda, ap, OUT_FILE):
waveform, alignment, spectrogram, mel_spectrogram, stop_tokens = synthesis(model, text, CONFIG, use_cuda, ap)
ap.save_wav(waveform, OUT_FILE)
return alignment, spectrogram, stop_tokens
def load_model(MODEL_PATH, sentence, CONFIG, use_cuda, OUT_FILE):
# load the model
num_chars = len(phonemes) if CONFIG.use_phonemes else len(symbols)
model = Tacotron(num_chars, CONFIG.embedding_size,['num_freq'],['num_mels'], CONFIG.r, attn_windowing=False)
# load the audio processor
#["power"] = 1.3["preemphasis"] = 0.97
ap = AudioProcessor(**
# load model state
if use_cuda:
cp = torch.load(MODEL_PATH)
cp = torch.load(MODEL_PATH, map_location=lambda storage, loc: storage)
# load the model
if use_cuda:
model.decoder.max_decoder_steps = 1000
align, spec, stop_tokens = tts(model, sentence, CONFIG, use_cuda, ap, OUT_FILE)
if __name__ == '__main__':
sentence = "Hello, how are you doing? My name is Sara"
load_model(MODEL_PATH, sentence, CONFIG, use_cuda, OUT_FILE)
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment