Speak Japanese text with OpenJTalk + python

Shows the method on Ubuntu and Windows. I'm sorry that Mac has no environment.

Talk with OpenJTalk + Python on Ubuntu 14.04 LTS

Open JTalk setup

Install open-jtalk (1.07) with ʻapt-get`.

Terminal

$ sudo apt-get install open-jtalk open-jtalk-mecab-naist-jdic hts-voice-nitech-jp-atr503-m001

The mecab dictionary is installed in/ var / lib / mecab / dic / open-jtalk / naist-jdic /.

Audio file setup

Voice data is diverted from MMDAgent. The MMDAgent_Example-1.6.zip is 11MB.

Terminal

$ wget https://sourceforge.net/projects/mmdagent/files/MMDAgent_Example/MMDAgent_Example-1.6/MMDAgent_Example-1.6.zip/download -O MMDAgent_Example-1.6.zip

Then extract the .htsvoice file

Terminal

$ unzip MMDAgent_Example-1.6.zip MMDAgent_Example-1.6/Voice/*

Copy the file to the same location as hts-voice-nitech

Terminal

$ sudo cp -r MMDAgent_Example-1.6/Voice/mei/ /usr/share/hts-voice

Playback with Python

http://raspi.seesaa.net/article/415530289.html I was allowed to refer to.

Save the following code as jtalk.py in a file.

jtalk.py

#coding: utf-8
import subprocess
from datetime import datetime

def jtalk(t):
    open_jtalk=['open_jtalk']
    mech=['-x','/var/lib/mecab/dic/open-jtalk/naist-jdic']
    htsvoice=['-m','/usr/share/hts-voice/mei/mei_normal.htsvoice']
    speed=['-r','1.0']
    outwav=['-ow','open_jtalk.wav']
    cmd=open_jtalk+mech+htsvoice+speed+outwav
    c = subprocess.Popen(cmd,stdin=subprocess.PIPE)
    c.stdin.write(t.encode())
    c.stdin.close()
    c.wait()
    aplay = ['aplay','-q','open_jtalk.wav']
    wr = subprocess.Popen(aplay)
    
def say_datetime():
    d = datetime.now()
    text = '%s month%s day,%s time%s minutes%s seconds' % (d.month, d.day, d.hour, d.minute, d.second)
    jtalk(text)

if __name__ == '__main__':
    say_datetime()

In the execution test

Terminal

$ python jtalk.py

Hopefully speak the current time When using from other modules

from_module_1.py

#coding: utf-8
import jtalk

jtalk.jtalk('Tell me something')

Set up Open JTalk on Windows

On Ubuntu, I could install it smoothly by doing ʻapt-get`, On Windows you build it yourself.

Obtaining the source code

Get OpenJTalk 1.09 from http://open-jtalk.sourceforge.net/

• http://downloads.sourceforge.net/open-jtalk/open_jtalk-1.09.tar.gz

Download the speech synthesis library hts_engineAPI required for compilation from http://hts-engine.sourceforge.net/

• http://downloads.sourceforge.net/hts-engine/hts_engine_API-1.10.tar.gz

If you expand ʻopen_jtalk-1.09.tar.gz to c: \ temp \ open_jtalk-1.09, then hts_engine_API-1.10.tar.gz is c: \ temp \ open_jtalk-1.09 \ hts_engine_API- Expanding to 1.10` will reduce problems when compiling.

Compile in Visual Studio

Launch the command tool in Visual Studio. Type nmake to make sure it works.

First, compile from hts_engine_API-1.10.

cmd

cd c:\temp\open_jtalk-1.09\hts_engine_API-1.10
nmake /f Makefile.mak
nmake /f Makefile.mak install

If it compiles successfully, a file will be created in c: \ hts_engine_API.

Next, compile open_jtalk.

cmd

cd c:\temp\open_jtalk-1.09
nmake /f Makefile.mak
nmake /f Makefile.mak install

When the character string to convert the dictionary is displayed near the end, the characters are garbled I'm curious, but if you're worried, Open JTalk has a prebuilt dictionary, so You can also use it.

This will generate c: \ open_jtalk \ bin \ open_jtalk.exe.

Get Voice data

From MMDAgent https://sourceforge.net/projects/mmdagent/files/MMDAgent_Example/MMDAgent_Example-1.6/ Download MMDAgent_Example-1.6.zip from, and copy Voice / mei / *. Htvoice to c: \ open_jtalk \ bin .

Execution test

Assuming that a file with Japanese input is created as input.txt in c: \ open_jtalk \ bin

cmd

c:\open\jtalk\bin
open_jtalk -m mei_normal.htsvoice -x ../dic -ow output.wav input.txt

I will try to execute. Now, if output.wav is created in the same folder, put it Double-click from Explorer to see if it plays.

Run OpenJTalk with Python on Windows

In ubuntu, I used the aplay command to play the file. On Windows Use the winsound module. Also, since the dictionary generation in this compilation on Windows was shift-jis, it is necessary to convert the encoding from the internal encoding of python when passing it to stdin.

Below, save the sample program as jtalk.py in c: \ open_jtalk \ bin.

jtalk.py

#coding: utf-8
# call OpenJTalk for windows
import subprocess
import winsound
from datetime import datetime


def jtalk(t):
    # depend on your install folder 
    OPENJTALK_BINPATH = 'c:/open_jtalk/bin'
    OPENJTALK_DICPATH = 'c:/open_jtalk/dic'
    OPENJTALK_VOICEPATH = 'c:/open_jtalk/bin/mei_normal.htsvoice'
    open_jtalk=[OPENJTALK_BINPATH + '/open_jtalk.exe']
    mech=['-x',OPENJTALK_DICPATH]
    htsvoice=['-m',OPENJTALK_VOICEPATH]
    speed=['-r','1.0']
    outwav=['-ow','open_jtalk.wav']
    cmd=open_jtalk+mech+htsvoice+speed+outwav
    c = subprocess.Popen(cmd,stdin=subprocess.PIPE)

    # convert text encoding from utf-8 to shitf-jis
    c.stdin.write(t.encode('shift-jis'))
    c.stdin.close()
    c.wait()

    # play wav audio file with winsound module
    winsound.PlaySound('open_jtalk.wav', winsound.SND_FILENAME)

    
def say_datetime():
    d = datetime.now()
    text = '%s month%s day,%s time%s minutes%s seconds' % (d.month, d.day, d.hour, d.minute, d.second)
    print(text)
    jtalk(text)

if __name__ == '__main__':
    say_datetime()

Do this. I have anaconda python installed, so

cmd

python jtalk.py

Then, speak the current date and time.

When using as a Python module

jtalk.py is saved in utf-8 encoding. When executing python from the command prompt, the Japanese character string is still Shift-JIS or is not unified.

Therefore, when entering Japanese, add ʻu" Japanese " and ʻu to indicate that it is utf-8.

from_module_2.py

import jtalk

jtalk.jtalk(u'I'll speak Japanese')