Get audio output directly

[jhoelzl]

Hello,

i have following code:

text = "Hello, my name is Josef."
targetLanguage = 'de'
tts = gTTS(text, targetLanguage)  
tts.save("synthesized.mp3")

Since i play the output directly after this code, is there a way to get the audio output directly as a variable? So i do not have to save and open the mp3 file again.

Thanks!
Josef

[pndurette]

Yes it's possible. You need a module that will play from a file-like/file object. Instead of doing tts.save() you can use tts.write_to_fp() to write to a file-like object that you can then reuse to play. As a really simple example:

from gtts import gTTS
from tempfile import TemporaryFile
tts = gTTS(text='Hello', lang='en')
f = TemporaryFile()
tts.write_to_fp(f)
# Play f
f.close()

[panev]

This would indeed be fantastic.

I'm having issues with finding a decent approach. Most audio playback libraries are hectic to use in this scanario.

[jhoelzl]

@pndurette thanks for help! Before, i used

os.system('cvlc --play-and-exit ' + "synthesized.mp3")

to play the audio file. Now, i have a an audio object called f.

I tried
os.system('cvlc --play-and-exit ' + f)

but is does not work. Do you have another example how to playback the generated audio object?

Thanks for helping!

Josef

[anselm94]

Try Pyglet - a multimedia and windowing library for Python (~3MB)

from gtts import gTTS
from time import sleep
import os
import pyglet

tts = gTTS(text='Hello World', lang='en')
filename = '/tmp/temp.mp3'
tts.save(filename)

music = pyglet.media.load(filename, streaming=False)
music.play()

sleep(music.duration) #prevent from killing
os.remove(filename) #remove temperory file

[jhoelzl]

Thanks @anselm94, but i want to avoid saving the audio data on the file system. Instead, i directly want to play the audio output from module gTTS. Do you think this is possible with pyglet?

[anselm94]

Currently Pyglet doesn't support using file-like objects. As the project documentation here (see Loading Media section) says

Future versions of pyglet will also support reading from arbitrary file-like objects, however a valid filename must currently be given.

[rishabhrkaushik]

@jhoelzl
Having the same issue. Have you found the solution yet.

[jhoelzl]

@rishabhrkaushik unfortunately not.

[Khoa-NT]

Hi, are there any ways to output file.wav , not mp3 ? i don't know how to use python play mp3 file.

The pyglet like anselm94 said can't open file .mp3

thank you and regards

[common83]

@shaolinkhoa using pyMedia lib
here example:
https://github.com/common83/playmp3/blob/master/tts_play.py

[AllenAkhaumere]

Use pygame multimedia package

[AllenAkhaumere]

pygame.mixer.init()
pygame.mixer.music.load(file.mp3)
pygame.mixer.music.play()
It works in real-time. Good luck

[yagiziskirik]

My solution to unicode-charactered languages:

#!/usr/bin/env python
#-*- coding: utf-8 -*-

from gtts import gTTS
import os

class Speech(object):
	def __init__(self):
		while True:
			try:
				a = raw_input("What you want me to say: ")
				a = unicode(a, "utf-8")
				tts = gTTS(text=a, lang="en")
				
				testfile = "/tmp/temp.mp3"
				tts.save(testfile)

				os.system("mpg123 /tmp/temp.mp3")
				os.system("clear")
				os.system("rm %s" %(testfile))
			except UnicodeDecodeError:
				print "Some characters are not supported."
spc = Speech()

You have to have a mpg123 to run this program, but it deletes the ".mp3" file everytime it finishes the phrase.

[VeNoMouS]

@yagiz190799

jesus... rm via os.system() really??? os.unlink() dude!

Also your clear can be safer with print "\033[H\033[J"

Might i also recommend pep8 standard, your using tabs for indentation (smh)... spaces mate!!

some people just want to see the world burn :P

[patel33hardik]

If are you using a web browser,
-save .mp3 to the temp folder on the server
-return file path or name

• use jQuery and HTML
  var file_name = response['Data']; var audio = new Audio(); audio.src = "frontend/temp_mp3/"+file_name; audio.addEventListener("loadedmetadata", function(_event) { var duration = (audio.duration * 1000); audio.play(); $("#TextToAudio").attr("src","frontend/images/text_to_audio_icon.gif"); deleteAudioFile(file_name) setTimeout(function(){ document.getElementById('TextToAudio').disabled=false; $("#TextToAudio").attr("src","frontend/images/text_to_audio_icon.png"); }, duration); });
• after content download, delete file from the server using ajax request.

[sorenmulli]

Hello everyone, I would like to share this solution using pyglet which was the only nice way I could resolve this without writing to disk:

import io
import time

import pyglet
from gtts import gTTS

pyglet.options["audio"] = ("pulse",)
TICK = .1

def speak(words: str, lang: str="en"):
    with io.BytesIO() as f:
        gTTS(text=words, lang=lang).write_to_fp(f)
        f.seek(0)
        sound = pyglet.media.load("_.mp3", file=f)
    player = sound.play()
    while player.playing:
        pyglet.app.platform_event_loop.dispatch_posted_events()
        pyglet.clock.tick()
        time.sleep(TICK)

Edit: Improved using suggestions by @VeNoMouS

[VeNoMouS]

Hello everyone, I would like to share this solution using pyglet which was the only nice way I could resolve this without writing to disk:

import io
import time

import pyglet
from gtts import gTTS

pyglet.options["audio"] = ("pulse",)
TICK = 0.2

def speak(words: str, lang: str="en"):
    with io.BytesIO() as f:
        gTTS(text=words, lang=lang).write_to_fp(f)
        f.seek(0)
        sound = pyglet.media.load("_.mp3", file=f)
    player = sound.play()
    playtime = player.source.duration
    while (playtime - TICK) > -
        pyglet.app.platform_event_loop.dispatch_posted_events()
        pyglet.clock.tick()
        playtime -= TICK
        time.sleep(TICK)

Would it not make more sense to do the following..

import io
import time

import pyglet
from gtts import gTTS

pyglet.options["audio"] = ("pulse",)

def speak(words: str, lang: str="en"):
    with io.BytesIO() as f:
        gTTS(text=words, lang=lang).write_to_fp(f)
        f.seek(0)
        
        player = pyglet.media.load('_.mp3', file=f).play()
        while player.source.duration != player.time:
            pyglet.app.platform_event_loop.dispatch_posted_events()
            pyglet.clock.tick()

Or better yet...

import io
import time

import pyglet
from gtts import gTTS

pyglet.options["audio"] = ("pulse",)

def speak(words: str, lang: str="en"):
    with io.BytesIO() as f:
        gTTS(text=words, lang=lang).write_to_fp(f)
        f.seek(0)
        
        player = pyglet.media.load('_.mp3', file=f).play()
        while player.playing:
            pyglet.app.platform_event_loop.dispatch_posted_events()
            pyglet.clock.tick()

[sorenmulli]

@VeNoMouS, that definitely looks nicer, but I can't get it to work: I only get the first small snippet of sound and only when i use dispatch_posted_events(), I get a satisfying result.

[VeNoMouS]

@sorenmulli updated those examples for you then ;P

[SG-SakethGupta]

Hey guys im trying to do a voice assistant

HELP

im trying to make the python file speak what i actually want it to speak. I have tried play sound and the temporary files. NO PIGLET or pygame plz

code:

def speak(text): speaker = gTTS(text = text, lang = "en") speaker.save("Voice.mp3") playsound.playsound("voice.mp3")

[themdrizwanansari]

@fasterastv2, do you have read access to the location where Voice.mp3 is being saved?

[SG-SakethGupta]

@themdrizwanansari Just realise that the problem isnt with playing sound.

The actual problem is with the saving part. The line speaker.save("Voice.mp3") causes the problem

[venzen]

Here is a solution to play gTTS sound directly using playsound and NamedTemporaryFile:

from gtts import gTTS
from tempfile import NamedTemporaryFile
from playsound import playsound

def speak(txt, lang='en'):
    gTTS(text=txt, lang=lang).write_to_fp(voice := NamedTemporaryFile())
    playsound(voice.name)
    voice.close()

txt = "Spank me, Daddy"
speak(txt)

Works in Python 3.9 and negates the need to save output to mp3.

[chajiuqqq]

@venzen ,thanks for your solution, only it works for me!

from gtts import gTTS
from tempfile import NamedTemporaryFile
from playsound import playsound

def speak(txt, lang='en'):
    gTTS(text=txt, lang=lang).write_to_fp(voice := NamedTemporaryFile())
    playsound(voice.name)
    voice.close()

txt = "Spank me, Daddy"
speak(txt)

[GrimPixel]

Here is a solution to play gTTS sound directly using playsound and NamedTemporaryFile:

from gtts import gTTS
from tempfile import NamedTemporaryFile
from playsound import playsound

def speak(txt, lang='en'):
    gTTS(text=txt, lang=lang).write_to_fp(voice := NamedTemporaryFile())
    playsound(voice.name)
    voice.close()

txt = "Spank me, Daddy"
speak(txt)

Works in Python 3.9 and negates the need to save output to mp3.

Ohh, my hands hurt.

[EyesOfSiva]

tank you @venzen and @GrimPixel that worked like a charm

[Vuizur]

I tried pyglet, pygame and the tempfile + playsound solution, but they all failed due to various bugs (Windows 10 Python 3.10). Audio play doesn't seem to be a very stable area, seriously, or my PC is broken.

Then I tried this solution:

from io import BytesIO
from pydub import AudioSegment
from pydub.playback import play

def say(text, lang = "en"):
    tts = gTTS(text=text, lang=lang)
    fp = BytesIO()
    tts.write_to_fp(fp)
    fp.seek(0)
    song = AudioSegment.from_file(fp, format="mp3")
    play(song)

which in theory should be efficient because it avoids writing the file on disk. Or so I thought, but it ended up being a few hundred ms slower than the naive solution which simply writes the file to disk. (I have an SSD though.) For a short text (5 words) the solution below is 200 - 300 ms faster.

from playsound import playsound

def speak_gtts(words: str, lang: str="en"):
    filename = datetime.now().strftime("%Y-%m-%d_%H-%M-%S") + ".mp3"
    tts = gTTS(words, lang=lang)
    tts.save(filename)
    playsound(filename)
    os.remove(filename)

[NinjaSkibidigyat]

from tempfile import NamedTemporaryFile
from playsound import playsound

def speak(txt, lang='en'):
    gTTS(text=txt, lang=lang).write_to_fp(voice := NamedTemporaryFile())
    playsound(voice.name)
    voice.close()

txt = "wow"
speak(txt)```

playsound is relying on another python subprocess. Please use `pip install pygobject` if you want playsound to run more efficiently.`
pip install pygobject
```Defaulting to user installation because normal site-packages is not writeable
Collecting pyobject
  Downloading pyobject-1.1.3.tar.gz (18 kB)
  Preparing metadata (setup.py) ... error
  error: subprocess-exited-with-error
  
  × python setup.py egg_info did not run successfully.
  │ exit code: 1
  ╰─> [8 lines of output]
      Traceback (most recent call last):
        File "<string>", line 2, in <module>
        File "<pip-setuptools-caller>", line 34, in <module>
        File "/tmp/pip-install-1xdtuxk1/pyobject_dddd7e8ca7a54e6ca5ba8e5b72577132/setup.py", line 8, in <module>
          long_desc=open("README.rst").read()
                    ^^^^^^^^^^^^^^^^^^^^^^^^^
        File "<frozen codecs>", line 322, in decode
      UnicodeDecodeError: 'utf-8' codec can't decode byte 0xb8 in position 13: invalid start byte
      [end of output]
  
  note: This error originates from a subprocess, and is likely not a problem with pip.
error: metadata-generation-failed

× Encountered error while generating package metadata.
╰─> See above for output.

note: This is an issue with the package mentioned above, not pip.
hint: See above for details.

[aadlinux358]

import io

from gtts import gTTS
from pydub import AudioSegment
from pydub.playback import play

hello = gTTS('hello')
fd = io.BytesIO()
hello.write_to_fp(fd)

fd.seek(0)

song = AudioSegment.from_mp3(fd)

play(song)

[BBC-Esq]

Loved the repository - thanks! Here's my contribution using sounddevice and soundfile:

SAMPLE SCRIPT

# pip install gTTS sounddevice soundfile numpy

from gtts import gTTS
import sounddevice as sd
import soundfile as sf
import io
from gtts.tokenizer import pre_processors, tokenizer_cases, Tokenizer

def preprocess_text(text):
    text = pre_processors.abbreviations(text)
    text = pre_processors.end_of_line(text)
    text = pre_processors.tone_marks(text)
    return text

def tokenize_and_minimize(text):
    tokenizer = Tokenizer([
        tokenizer_cases.tone_marks,
        tokenizer_cases.period_comma,
        tokenizer_cases.colon,
        tokenizer_cases.other_punctuation
    ])
    
    tokens = tokenizer.run(text)
    
    minimized_tokens = []
    for token in tokens:
        if not token.strip():
            continue
        while len(token) > 100:
            split_index = token.rfind(' ', 0, 100)
            if split_index == -1:
                split_index = 100
            minimized_tokens.append(token[:split_index])
            token = token[split_index:].lstrip()
        minimized_tokens.append(token)
    
    return minimized_tokens

text = "Hello, how are you doing today? I'm doing well thank you for asking. Dr. Smith Jr. will see you now. This is a very-long-hyphenated-word. Let's add some tone marks: Wow! Amazing? Great!"

processed_text = preprocess_text(text)
print("Processed text:", processed_text)

tokens = tokenize_and_minimize(processed_text)
print("Tokens:", tokens)

fp = io.BytesIO()

# Process each token and concatenate the audio
for token in tokens:
    if token.strip():
        print(f"Processing token: '{token}'")
        # Default: English
        tts = gTTS(text=token, lang='en', slow=False, tld='com.ng')
        # For Spanish:
        # tts = gTTS(text=token, lang='es', slow=False)
        # For French:
        # tts = gTTS(text=token, lang='fr', slow=False)
        # For German:
        # tts = gTTS(text=token, lang='de', slow=False)
        # For Japanese:
        # tts = gTTS(text=token, lang='ja', slow=False)
        tts.write_to_fp(fp)

fp.seek(0)

# Read the audio data using soundfile
data, samplerate = sf.read(fp)

# Play the audio
sd.play(data, samplerate)
sd.wait()