Recording synthesized text-to-speech to a file in Python

Question

I am attempting to find a way to take synthesized speech and record it to an audio file. I am currently using pyttsx as my text-to-speech library, but there isn't a mechanism for saving the output to a file, only playing it directly from the speakers. I've looked into detecting and recording audio as well as PyAudio, but these seem to take input from a microphone rather than redirecting outgoing audio to a file. Is there a known way to do this?

Answer 1

You can call espeak with the -w argument using subprocess.

import subprocess

def textToWav(text,file_name):
   subprocess.call(["espeak", "-w"+file_name+".wav", text])

textToWav('hello world','hello')

This will write file_name.wav without reading out loud. If your text is in a file (e.g. text.txt) you need to call espeak with the -f parameter ("-f"+text). I'd recommend reading the espeak man pages to see all the options you have.

Hope this helps.

Answer 2

You can use more advanced SAPI wrapper to save output to the wav file. For example you can try

https://github.com/DeepHorizons/tts

The code should look like this:

import tts.sapi
voice = tts.sapi.Sapi()
voice.set_voice("Joey")
voice.create_recording('hello.wav', "Hello")

Answer 3

Here is an example which gives you access to the NSSpeechSynthesizer API

#!/usr/bin/env python

from  AppKit import NSSpeechSynthesizer
import sys
import Foundation


if len(sys.argv) < 2:
   text = raw_input('type text to speak> ')
else:
   text = sys.argv[1]

nssp = NSSpeechSynthesizer
ve = nssp.alloc().init()
ve.setRate_(100)
url = Foundation.NSURL.fileURLWithPath_('yourpath/test.aiff')
ve.startSpeakingString_toURL_(text,url)

Answer 4

use a VB in and out emulator cable with audacity to record whatever engine.play() outputs.

You will need to download the VB Cables, then set them up through windows as default with windows sound settings,then set Audacity's input to Microsoft sound mapper input, hit record on Audacity and play your script. You will notice once's it's finished on Audacity and then export what you need to loop.

This method may be out of the scope is not perfect but works on pyttsx3 and records in better quality than espeak.

Answer 5

You can use Amazon Polly API as well:

An example using python is available below:

https://aws.amazon.com/blogs/machine-learning/convert-your-text-into-an-mp3-file-with-amazon-polly-and-a-simple-python-script/