Watson Speech to Text Conversion

convert.rb

#!/usr/bin/env ruby

require 'json'
require 'net/http'
require 'dotenv/load'

TYPE_OF_FILE_TO_CONVERT = '*.mp4'

# Sign up for an account and confirm your email
# https://console.bluemix.net/registration/

# Generate a Speech to Text App
# https://console.bluemix.net/catalog/services/speech-to-text

# Copy the API Key and the Service URL to these Environment Variables
WATSON_API_KEY = ENV['WATSON_API_KEY']
WATSON_SERVICE_URL = ENV['WATSON_SERVICE_URL']
# i.e., for DC - WATSON_SERVICE_URL = 'https://gateway-wdc.watsonplatform.net/speech-to-text/api/v1/recognize'

# Install some tools first if you need to convert video to mp3
# brew install ffmpeg
# brew install lame

# Place this file in the directory that you have files to convert.
# Make this file executable
# chmod +x convert.rb

# Let er rip!
# ./convert.rb

def convert
  Dir.glob(TYPE_OF_FILE_TO_CONVERT) do |file|
    @file = file

    # This step can be commented out if your file is already an MP3
    convert_file_to_mp3(file)
    convert_mp3_to_text
  end
end

def convert_file_to_mp3(input)
  `ffmpeg -i #{input} -b:a 320K -vn #{mp3_output} -y`
end

def convert_mp3_to_text
  generate_text_file(make_api_request)
end

def make_api_request
  uri = URI.parse(WATSON_SERVICE_URL)
  request = Net::HTTP::Post.new(uri)
  request.basic_auth('apikey', WATSON_API_KEY)
  request.content_type = 'audio/mp3'
  request.body = File.read(mp3_output)
  req_options = { use_ssl: uri.scheme == 'https' }
  Net::HTTP.start(uri.hostname, uri.port, req_options) do |http|
    http.request(request)
  end
end

def generate_text_file(response)
  body = response.body
  json = JSON.parse(body)

  text = json.dig('results').to_a.map { |e| e.dig('alternatives') }.to_a.flatten.map { |d| d.dig('transcript') }.to_a.join(' ').gsub('  ', ' ')
  File.open(text_output, 'w') do |file|
    file.write text
  end
end

def basename
  File.basename(@file, '.*')
end

def mp3_output
  basename + '.mp3'
end

def text_output
  basename + '.txt'
end

convert