Text to phonemes.ipynb

text-to-phonemes.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Text to phonemes.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "authorship_tag": "ABX9TyPxt7g1EwF5xXLqbF28l2ha",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/gist/NTT123/ec538440bb5ab5a4209d10d1cd25e105/text-to-phonemes.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "- We use `phonemizer==3.1.1` library with `espeak-ng v1.51` backend.\n",
        "    + Fixed the versions to make sure that the output is deterministic.\n",
        "- Reuse the espeak backend in the python example to speedup the converting."
      ],
      "metadata": {
        "id": "YDowu3Hy57Ms"
      }
    },
    {
      "cell_type": "code",
      "execution_count": 1,
      "metadata": {
        "id": "RoC5hBxiUad0"
      },
      "outputs": [],
      "source": [
        "# install espeak-ng 1.51 and phonemizer 3.1.1\n",
        "%%capture\n",
        "!rm -rf espeak\n",
        "!mkdir -p espeak\n",
        "!cd espeak; wget https://github.com/espeak-ng/espeak-ng/archive/refs/tags/1.51.zip\n",
        "!cd espeak; unzip -qq 1.51.zip\n",
        "!sudo apt install -y make autoconf automake libtool pkg-config gcc\n",
        "!cd espeak/espeak-ng-1.51; ./autogen.sh; ./configure --prefix=`pwd`/../usr\n",
        "!cd espeak/espeak-ng-1.51; make; make install\n",
        "!pip install -U pip\n",
        "!pip install -U phonemizer==3.1.1"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# let phonemizer know where to load espeak\n",
        "import os\n",
        "os.environ[\"PHONEMIZER_ESPEAK_LIBRARY\"] = \"./espeak/usr/lib/libespeak-ng.so.1.1.51\""
      ],
      "metadata": {
        "id": "nhaU-G054sQk"
      },
      "execution_count": 2,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# command-line example\n",
        "!echo \"Modern text-to-speech synthesis pipelines typically involve multiple processing stages.\" | phonemize -l en-us -b espeak --with-stress --strip --preserve-punctuation"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "KKZKoGK-U6w2",
        "outputId": "cc1e5410-5ae2-4e83-d3a5-5a317919ebeb"
      },
      "execution_count": 3,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "mˈɑːdɚn tˈɛksttəspˈiːtʃ sˈɪnθəsˌɪs pˈaɪplaɪnz tˈɪpɪkli ɪnvˈɑːlv mˌʌltɪpəl pɹˈɑːsɛsɪŋ stˈeɪdʒᵻz.\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# python example\n",
        "%%time\n",
        "from phonemizer.backend import EspeakBackend\n",
        "backend = EspeakBackend('en-us', preserve_punctuation=True, with_stress=True)\n",
        "s = \"Modern text-to-speech synthesis pipelines typically involve multiple processing stages.\" \n",
        "a = []\n",
        "for i in range(10000):\n",
        "    p = backend.phonemize([s], strip=True)\n",
        "    a.append(p[0])\n",
        "print(a[0])"
      ],
      "metadata": {
        "id": "5cN6l0VLVPJF",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "9b9cd956-eb20-4890-fabe-e0d11d637ea7"
      },
      "execution_count": 4,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "mˈɑːdɚn tˈɛksttəspˈiːtʃ sˈɪnθəsˌɪs pˈaɪplaɪnz tˈɪpɪkli ɪnvˈɑːlv mˌʌltɪpəl pɹˈɑːsɛsɪŋ stˈeɪdʒᵻz.\n",
            "CPU times: user 6.82 s, sys: 55.9 ms, total: 6.87 s\n",
            "Wall time: 7.1 s\n"
          ]
        }
      ]
    }
  ]
}