the code is mostly from pyttsx3.
only because the repo pyttsx3 does not update for years and some new feature i want is not here, i cloned this repo.
feature:
1 nsss
2 sapi5
3 espeak
4 coqui_ai_tts
1 say
engine = pyttsx4.init()
engine.say('this is an english text to voice test.')
engine.runAndWait()
2 save to file
import pyttsx4
engine = pyttsx4.init()
engine.save_to_file('i am Hello World, i am a programmer. i think life is short.', 'test1.wav')
engine.runAndWait()
1 memory support for sapi5, nsss, espeak. NOTE: the memory is just raw adc data, wav header has to be added if you want to save to wav file.
import pyttsx4
from io import BytesIO
from pydub import AudioSegment
from pydub.playback import play
import os
import sys
engine = pyttsx4.init()
b = BytesIO()
engine.save_to_file('i am Hello World', b)
engine.runAndWait()
#the bs is raw data of the audio.
bs=b.getvalue()
# add an wav file format header
b=bytes(b'RIFF')+ (len(bs)+38).to_bytes(4, byteorder='little')+b'WAVEfmt\x20\x12\x00\x00' \
b'\x00\x01\x00\x01\x00' \
b'\x22\x56\x00\x00\x44\xac\x00\x00' +\
b'\x02\x00\x10\x00\x00\x00data' +(len(bs)).to_bytes(4, byteorder='little')+bs
# changed to BytesIO
b=BytesIO(b)
audio = AudioSegment.from_file(b, format="wav")
play(audio)
sys.exit(0)
2 cloning voice
# only coqui_ai_tts engine support cloning voice.
engine = pyttsx4.init('coqui_ai_tts')
engine.setProperty('speaker_wav', './docs/i_have_a_dream_10s.wav')
engine.say('this is an english text to voice test, listen it carefully and tell who i am.')
engine.runAndWait()
voice clone test1:
voice clone test2:
the changelog:
- add memory support for sapi5
- add memory support for espeak(espeak is not tested). eg:
b = BytesIO()
engine.save_to_file('i am Hello World', b)
engine.runAndWait()
-
fix VoiceAge key error
-
fix for sapi save_to_file when it run on machine without outputsream device.
-
fix save_to_file does not work on mac os ventura error. --3.0.6
-
add pitch support for sapi5(not tested yet). --3.0.8
-
fix nsss engine: Import super from objc to fix AttributeError by @matt-oakes.
-
add tts support: deep-learning text to voice backend:
just say:
engine = pyttsx4.init('coqui_ai_tts')
engine.say('this is an english text to voice test.')
engine.runAndWait()
cloning someones voice:
engine = pyttsx4.init('coqui_ai_tts')
engine.setProperty('speaker_wav', './someones_voice.wav')
engine.say('this is an english text to voice test.')
engine.runAndWait()
demo output:
NOTE:
if save_to_file with BytesIO, there is no wav header in the BytesIO. the format of the bytes data is that 2-bytes = one sample.
if you want to add a header, the format of the data is: 1-channel. 2-bytes of sample width. 22050-framerate.
how to add a wav header in memory:#2
install:
pip install pyttsx4
use:
import pyttsx4
engine = pyttsx4.init()
the other usage is the same as the pyttsx3
https://pyttsx3.readthedocs.io/en/latest/
- sapi5
- nsss
- espeak
Feel free to wrap another text-to-speech engine for use with pyttsx4.