Example code of ruby with Amazon Polly

Gemfile

source 'https://rubygems.org'

gem 'nokogiri', '~>1.6'
gem 'aws-sdk', '~> 2'

polly_synthesizer.rb

require 'aws-sdk'
require 'nokogiri'
require 'open-uri'

class Synthesizer
  def initialize(region='us-east-1')
    @polly = Aws::Polly::Client.new(region: region)
  end

  def synthesize(text, file_name="./tmp.mp3", voice_id="Joanna")
    @polly.synthesize_speech(
      response_target: file_name,
      text: text,
      output_format: "mp3",
      # You can use voice IDs http://docs.aws.amazon.com/polly/latest/dg/API_Voice.html
      # If you want to synthesize Japanese voice, you can use "Mizuki"
      voice_id: voice_id
    )
  end
end

module TextFetcher
  def self.fetch_text_from(url, xpath)
    charset = nil
    html = open(url) do |f|
      charset = f.charset
      f.read
    end
    doc = Nokogiri::HTML.parse(html, nil, charset)
    node_texts = doc.xpath(xpath).map(&:text)

    combined_texts = []
    tmp_string = ""
    node_texts.each do |text|
      if tmp_string.size + text.size > 1500
        combined_texts << tmp_string
        tmp_string = text
      else
        tmp_string << " #{text}"
      end
    end
    combined_texts << tmp_string
  end
end

if __FILE__ == $0
  synthesizer = Synthesizer.new

  url = "https://medium.com/@chezou/building-predictive-model-with-ibis-impala-and-scikit-learn-356b41f404e0#.xeiwrmhhb"
  # This XPath assumes medium contents
  xpath = '//section//text()'
  input_texts = TextFetcher.fetch_text_from(url, xpath)

  input_texts.each.with_index do |text, i|
    synthesizer.synthesize(text, "./tmp_#{i}.mp3")
    sleep(1)
  end
  # You can combine mp3 with cat on Linux based system
  `cat ./tmp_*.mp3 > ./combined.mp3`
end

Thanks a lot for this. Have used the snippets and I think in the do iteration over the array where you construct the 1500 bytes chunks it has an error. In my case it repeats certain sentences. Can you please review?

ok, found the duplication issue with help from stackoverflow: Correct is.

node_texts.each do |text|
      if tmp_string.size + text.size > 1500
        combined_texts << tmp_string
        tmp_string = ""
      end
      tmp_string << " #{text}"
    end
    combined_texts << tmp_string
  end

or even better:

if tmp_string.size + text.size > 1500
  combined_texts << tmp_string
  tmp_string = text
else
  tmp_string << " #{text}"
end

@mkuendig Thanks for your patch! I fixed it :)

I created ruby gem https://rubygems.org/gems/pollynomial