Team Project Overview

• https://github.com/brainmining-for-bigdata/project-text2speech-blog

• Text2Speech를 구현한 Django Project

• Model : tacotron 사용

Project 실행 방법

1. git clone
2. Database생성- user: root pw: 1234

• database생성시 반드시 character set을 지정해야 함

mysql> create database maratron character set utf8
mysql> show databases;

3. default table, tts App 내 DB 모델 생성 및 반영

python manage.py migrate
반영 여부 확인
python manage.py showmigrations tts
강제로 migrate할 때 
python manage.py migrate --fake tts zero

superUser 계정 생성
python manage.py createsuperuser
서버 구동하여 admin 페이지 정상 접속 여부 확인
python manage.py runserver

4. Python manage.py run
5. web browser 접속
   http://127.0.0.1:8000/admin/

commands

Preprocess the data

python preprocess.py --dataset ljspeech

Train a model

python train.py

Continue training from the the checkpoint

python train.py --restore_step=9000

Monitor with Tensorboard

tensorboard --logdir ~/tacotron/logs-tacotron

Evaluate

python eval.py --checkpoint ~/tacotron/logs-tacotron/model.ckpt-185000

Make audio book

python eval.py --checkpoint ./logs-tacotron-model/model.ckpt-64000 --cleaners korean_cleaners --audiobook ./audiobooks/이기적유전자.txt python eval.py --checkpoint ./LJlogs-tacotron/model.ckpt-40000 --audiobook ./audiobooks/Secret.txt

add audiobook into database

python initdata.py

Run a server

python manage.py runserver