tts.py

"""
Synthesizes speech from the input string of text.
from reader.services.translate.polly import text_to_speech_mp3
reading = Reading.objects.last()
with open('./test.mp3', 'wb') as f:
    text_to_speech_mp3(reading, f)
"""
from io import BytesIO

import boto3
from pydub import AudioSegment

from .transcript import get_sample_parts, get_transcript_parts

polly = boto3.Session(region_name='ap-southeast-2').client('polly')

VOICE_NAME = 'Matthew'
MAX_REQUEST_LENGTH = 2500

def text_to_speech_mp3(reading, out_file):
    paragraphs = get_transcript_parts(reading)
    mp3_parts = get_mp3_parts(paragraphs)
    write_reading_audio(mp3_parts, out_file)


def write_reading_audio(mp3_parts, out_file):
    speech_audio = None
    half_sec_pause = AudioSegment.silent(duration=500)
    for mp3_part in mp3_parts:
        audio_segment = AudioSegment.from_mp3(mp3_part)
        if speech_audio:
            speech_audio += audio_segment
        else:
            speech_audio = audio_segment

        # Add a 0.5s pause after a paragraph
        speech_audio += half_sec_pause

    speech_audio.export(out_file, format='mp3')


def get_mp3_parts(paragraphs):
    # TODO - multi thread this to get better performance
    parts = []
    for counter, paragraph in enumerate(paragraphs):
        text_batches = batch(paragraph, MAX_REQUEST_LENGTH)
        for text_batch in text_batches:
            parts.append(text_to_mp3_bytes(text_batch))

    return parts


def text_to_mp3_bytes(text):
    assert len(text) <= 5000
    response = polly.synthesize_speech(
        VoiceId=VOICE_NAME,
        OutputFormat='mp3', 
        Text =text
    )
    return BytesIO(response['AudioStream'].read())


def batch(iterable, n=1):
    l = len(iterable)
    for ndx in range(0, l, n):
        yield iterable[ndx : min(ndx + n, l)]