How can I efficiently convert gtts audio into pydub audiosegments?

Question

I want to manipulate gtts audio in pydub but I am not sure how to translate gtts file like to pydub audio.

I know that I can convert google text to speech audio to an mp3 and I know that I can import an mp3 with pydub, but this process is inefficient. Is there a way to skip creating mp3 file and reading it back in?

I tried using the gtts file-like as the argument in AudioSegment.from_mp3() but I am pretty sure it is looking for a string.

from gtts import gTTS
from io import BytesIO
from pydub import AudioSegment

mp3_fp = BytesIO()
tts = gTTS('hello', 'en')
tts.write_to_fp(mp3_fp)
song = AudioSegment.from_mp3(mp3_fp)

I got a "CouldntDecodeError":

  File "C:\ProgramData\Anaconda3\lib\site-packages\spyder\utils\site\sitecustomize.py", line 102, in execfile
    exec(compile(f.read(), filename, 'exec'), namespace)

  File "C:/Users/py/ex/gtts_test.py", line 18, in <module>
    song = AudioSegment.from_mp3(mp3_fp)

  File "C:\ProgramData\Anaconda3\lib\site-packages\pydub\audio_segment.py", line 716, in from_mp3
    return cls.from_file(file, 'mp3', parameters=parameters)

  File "C:\ProgramData\Anaconda3\lib\site-packages\pydub\audio_segment.py", line 704, in from_file
    p.returncode, p_err))

CouldntDecodeError: Decoding failed. ffmpeg returned error code: 1

Output from ffmpeg/avlib:

b'ffmpeg version 4.1.1 Copyright (c) 2000-2019 the FFmpeg developers\r\n  built with gcc 8.2.1 (GCC) 20190212\r\n  configuration: --enable-gpl --enable-version3 --enable-sdl2 --enable-fontconfig --enable-gnutls --enable-iconv --enable-libass --enable-libbluray --enable-libfreetype --enable-libmp3lame --enable-libopencore-amrnb --enable-libopencore-amrwb --enable-libopenjpeg --enable-libopus --enable-libshine --enable-libsnappy --enable-libsoxr --enable-libtheora --enable-libtwolame --enable-libvpx --enable-libwavpack --enable-libwebp --enable-libx264 --enable-libx265 --enable-libxml2 --enable-libzimg --enable-lzma --enable-zlib --enable-gmp --enable-libvidstab --enable-libvorbis --enable-libvo-amrwbenc --enable-libmysofa --enable-libspeex --enable-libxvid --enable-libaom --enable-libmfx --enable-amf --enable-ffnvcodec --enable-cuvid --enable-d3d11va --enable-nvenc --enable-nvdec --enable-dxva2 --enable-avisynth\r\n  libavutil      56. 22.100 / 56. 22.100\r\n  libavcodec     58. 35.100 / 58. 35.100\r\n  libavformat    58. 20.100 / 58. 20.100\r\n  libavdevice    58.  5.100 / 58.  5.100\r\n  libavfilter     7. 40.101 /  7. 40.101\r\n  libswscale      5.  3.100 /  5.  3.100\r\n  libswresample   3.  3.100 /  3.  3.100\r\n  libpostproc    55.  3.100 / 55.  3.100\r\n[mp3 @ 000001da0c1292c0] Failed to read frame size: Could not seek to 1026.\r\npipe:: Invalid argument\r\n'

Answer 1

You must seek the BytesIO object to 0!

from gtts import gTTS
from io import BytesIO
from pydub import AudioSegment

mp3_fp = BytesIO()
tts = gTTS('hello', 'en')
tts.write_to_fp(mp3_fp)
mp3_fp.seek(0)
song = AudioSegment.from_mp3(mp3_fp)

Answer 2

if you have very long text like wikipeda summary it can take too long to convert it to the sound. the script below is taking long text and splitting to the sentences and converting them to sound and adding one mp3 file also i added funny part to change sound speed and make it like child voice.have fun...

from pydub import AudioSegment
from gtts import gTTS
from io import BytesIO
import pygame
from nltk.tokenize import sent_tokenize
from queue import Queue  # Python 3 import
import threading

sentence_queue = Queue()
audio_queue = Queue()


def split_sentence(text):
    # text = text.replace('.', '. ')
    text = text.replace('!', '. ')
    text = text.replace('?', '. ')
    sentences = sent_tokenize(text)
    for sentence in sentences:
        sentence_queue.put(sentence)
    sentence_queue.put(None)


def speed_swifter(sound, speed=1.0):
    sound_with_altered_frame_rate = sound._spawn(
        sound.raw_data, overrides={"frame_rate": int(sound.frame_rate * speed)})
    return sound_with_altered_frame_rate


def recognize_worker():
    # this runs in a background thread
    i = 1
    while True:
        mp3_fp = BytesIO()
        # retrieve the next sentense job from the main thread
        sentence = sentence_queue.get()
        if sentence is None:
            break
        # received sentence data, now we'll convert it using Google
        try:
            speech = gTTS(text=sentence, lang="tr", slow=False)
            speech.write_to_fp(mp3_fp)  # save to IO bytes
        except:
            print("gtts çal??mad?")
        sentence_queue.task_done()  # mark the audio processing job as completed in the queue
        mp3_fp.seek(0)  # BYTE IO rewind ?
        song = AudioSegment.from_mp3(mp3_fp)  # convert to audio
        slower_sound = speed_swifter(song, 1.3)
        slower_sound.export('assets/sound_out/'+str(i)+'.mp3', format="mp3")
        audio_queue.put('assets/sound_out/'+str(i)+'.mp3')
        i += 1
    audio_queue.put(None)


def player(filename):
    pygame.mixer.init()
    pygame.mixer.music.load(filename)
    pygame.mixer.music.set_volume(1.0)
    pygame.mixer.music.play()
    while pygame.mixer.music.get_busy() == True:
        pass
    pygame.mixer.music.unload()
    os.remove(filename)


def say(context):
    split_sentence(context)
    threading.Thread(target=recognize_worker).start()
    while True:
        mp3 = audio_queue.get()
        if mp3 is None:
            break
        player(mp3)
        audio_queue.task_done()


if __name__ == "__main__":
    context = 'Murat Boz. Türk ?ark?c?. söz yazar?. ve oyuncudur.ürün bekledi?imizden. çok çok iyi. gerek görüntü. kalitesi gerek de ses kalitesi gayet güzel. dü?ünmeden alabilirsiniz '
    say(context)
``