音声合成マークアップ言語(SSML)の利用例2-英語による「AIに関するピーターパンとアラレちゃんの仮想対話」

ここでは、Google Text-to-Speechにおいて、音声合成マークアップ言語(Speech Synthesis Markup Language;SSML)を用いた発話者の指定法についてのサンプルを挙げる。

  1. GoogleのText-to-Speechを利用して作成したAI音声
    サンプルSSMLファイル例1にあるように、ナレーション部はen-GB-Neural2-D、Peter Panはen-GB-Neural2-B、アラレちゃんはen-GB-Neural2-Cを利用したバージョン。
     

     
    2024年5月14日現在では、Google Text-to-Speechのイギリス英語化におけるWEBでの対話的利用においてWavenetおよびNews Voicesの音声を使用することは、下記のようなエラー表示が出てできない。そのため下記サンプルでは、Neral2音声を利用している。

    Error synthesizing speech
    The requested voice cannot support ssml tags that use Wavenet or News voices. Please modify your request and try again.
     

  2. GoogleのText-to-Speechを利用して作成したアメリカ英語版AI音声
    サンプルSSMLファイル例2にあるように、ナレーション部はen-US-Wavenet-I、Peter Panはen-US-Wavenet-D、アラレちゃんはen-US-Wavenet-Gを利用したバージョン。

    Google Text-to-Speechのアメリカ英語化におけるWEBでの対話的利用においてはイギリス英語音声とは逆にNeural2音声が利用できない。下記のようなエラー表示にあるように、アメリカ英語化におけるWEBでの対話的利用においてはNeural2音声とStudio音声が利用できない。そのため下記サンプルでは、Wavenet音声を利用している。

     
    The requested voice cannot support ssml tags that use Neural2 or Studio voices. Please modify your request and try again.
     
  3. 比較参考のために、ChatGPT4oの音声で読み上げた音声ファイルを下記に挙げる。ChatGPT4oではSSML言語のようなきめ細やかな指示による音声読み上げはできないが、GoogleのText-to-Speechよりも、音量変化なども含め、より自然な読み上げになっている。
     
    1. 読み上げ音声:Cove

    2. 読み上げ音声:Juniper
     
    --- 以下、サンプルSSMLファイル例1---

    <div class=”quote”>
    <speak>
    <voice language=”en-GB” gender=”male” required=”gender” ordering=”gender language” name=”en-GB-Neural2-D”>
    Peter Pan and Arale-chan’s Discussion about AI
    <break time=”2s”/>
    </voice>

    <voice language=”en-GB” gender=”male” required=”gender” ordering=”gender language” name=”en-GB-Neural2-B”>Peter Pan: <break time=”900ms”/>Arale-chan, have you heard of AI?</voice>
    <voice language=”en-GB” gender=”female” required=”gender” ordering=”gender language” name=”en-GB-Neural2-C”>Arale: <break time=”900ms”/>Yes, I know! AI stands for artificial intelligence. Is it something like me, created by Senbei Norimaki in ‘Dr. Slump’?</voice>
    <voice language=”en-GB” gender=”male” required=”gender” ordering=”gender language” name=”en-GB-Neural2-B”>Peter Pan: <break time=”900ms”/>Yes, that’s right. But AI is more than just moving robots. it can also talk like you.</voice>
    <voice language=”en-GB” gender=”female” required=”gender” ordering=”gender language” name=”en-GB-Neural2-C”>Arale: <break time=”900ms”/>Wow, that’s interesting! What kinds of things can AI talk about?</voice>
    <voice language=”en-GB” gender=”male” required=”gender” ordering=”gender language” name=”en-GB-Neural2-B”>Peter Pan: <break time=”900ms”/>Various topics. AI learns human languages and can converse like us. However, AI speaks based on programmed information, which means it doesn’t actually have feelings.</voice>
    <voice language=”en-GB” gender=”female” required=”gender” ordering=”gender language” name=”en-GB-Neural2-C”>Arale: <break time=”900ms”/>I see. I’m programmed too, yet I find many things fun and interesting!</voice>
    <voice language=”en-GB” gender=”male” required=”gender” ordering=”gender language” name=”en-GB-Neural2-B”>Peter Pan: <break time=”900ms”/>That’s true. What do you think about AI assisting or replacing the work of human writers and artists?</voice>
    <voice language=”en-GB” gender=”female” required=”gender” ordering=”gender language” name=”en-GB-Neural2-C”>Arale: <break time=”900ms”/>Hmm, it’s impressive that AI can draw and write stories, but it might lack something essential without a human artist’s heart and soul.</voice>
    <voice language=”en-GB” gender=”male” required=”gender” ordering=”gender language” name=”en-GB-Neural2-B”>Peter Pan: <break time=”900ms”/>Yes, AI is highly intelligent, but it cannot feel emotions like humans, which raises questions about the soulfulness of its work.</voice>
    <voice language=”en-GB” gender=”female” required=”gender” ordering=”gender language” name=”en-GB-Neural2-C”>Arale: <break time=”900ms”/>However, it would be wonderful if AI could collaborate with humans to create even more beautiful things!</voice>
    <voice language=”en-GB” gender=”male” required=”gender” ordering=”gender language” name=”en-GB-Neural2-B”>Peter Pan: <break time=”900ms”/>That’s an excellent idea, which could lead to new possibilities for collaboration between AI and humans.</voice>
    <voice language=”en-GB” gender=”female” required=”gender” ordering=”gender language” name=”en-GB-Neural2-C”>Arale: <break time=”900ms”/>Yes, that’s right! Let’s team up and have lots of fun adventures!</voice>

    <voice language=”en-GB” gender=”male” required=”gender” ordering=”gender language” name=”en-GB-Neural2-D”>
    In this dialogue, Peter Pan and Arale-chan engage in a friendly discussion about AI’s capabilities and limitations. They exchange views on how AI could potentially assist or replace human creative work and ultimately recognise the importance of AI and humans working together.
    </voice>
    </speak>
    </div>

    --- 以下、サンプルSSMLファイル例2---

    <div class=”quote”>
    <speak>
    <voice language=”en-US” gender=”male” required=”gender” ordering=”gender language” name=”en-US-Wavenet-I”>
    Peter Pan and Arale-chan’s Discussion about AI
    <break time=”2s”/>
    </voice>

    <voice language=”en-US” gender=”male” required=”gender” ordering=”gender language” name=”en-US-Wavenet-D”>Peter Pan: <break time=”900ms”/>Arale-chan, have you heard of AI?</voice>
    <voice language=”en-US” gender=”female” required=”gender” ordering=”gender language” name=”en-US-Wavenet-G”>Arale: <break time=”900ms”/>Yes, I know! AI stands for artificial intelligence. Is it something like me, created by Senbei Norimaki in ‘Dr. Slump’?</voice>
    <voice language=”en-US” gender=”male” required=”gender” ordering=”gender language” name=”en-US-Wavenet-D”>Peter Pan: <break time=”900ms”/>Yes, that’s right. But AI is more than just moving robots. it can also talk like you.</voice>
    <voice language=”en-US” gender=”female” required=”gender” ordering=”gender language” name=”en-US-Wavenet-G”>Arale: <break time=”900ms”/>Wow, that’s interesting! What kinds of things can AI talk about?</voice>
    <voice language=”en-US” gender=”male” required=”gender” ordering=”gender language” name=”en-US-Wavenet-D”>Peter Pan: <break time=”900ms”/>Various topics. AI learns human languages and can converse like us. However, AI speaks based on programmed information, which means it doesn’t actually have feelings.</voice>
    <voice language=”en-US” gender=”female” required=”gender” ordering=”gender language” name=”en-US-Wavenet-G”>Arale: <break time=”900ms”/>I see. I’m programmed too, yet I find many things fun and interesting!</voice>
    <voice language=”en-US” gender=”male” required=”gender” ordering=”gender language” name=”en-US-Wavenet-D”>Peter Pan: <break time=”900ms”/>That’s true. What do you think about AI assisting or replacing the work of human writers and artists?</voice>
    <voice language=”en-US” gender=”female” required=”gender” ordering=”gender language” name=”en-US-Wavenet-G”>Arale: <break time=”900ms”/>Hmm, it’s impressive that AI can draw and write stories, but it might lack something essential without a human artist’s heart and soul.</voice>
    <voice language=”en-US” gender=”male” required=”gender” ordering=”gender language” name=”en-US-Wavenet-D”>Peter Pan: <break time=”900ms”/>Yes, AI is highly intelligent, but it cannot feel emotions like humans, which raises questions about the soulfulness of its work.</voice>
    <voice language=”en-US” gender=”female” required=”gender” ordering=”gender language” name=”en-US-Wavenet-G”>Arale: <break time=”900ms”/>However, it would be wonderful if AI could collaborate with humans to create even more beautiful things!</voice>
    <voice language=”en-US” gender=”male” required=”gender” ordering=”gender language” name=”en-US-Wavenet-D”>Peter Pan: <break time=”900ms”/>That’s an excellent idea, which could lead to new possibilities for collaboration between AI and humans.</voice>
    <voice language=”en-US” gender=”female” required=”gender” ordering=”gender language” name=”en-US-Wavenet-G”>Arale: <break time=”900ms”/>Yes, that’s right! Let’s team up and have lots of fun adventures!</voice>

    <voice language=”en-US” gender=”male” required=”gender” ordering=”gender language” name=”en-US-Wavenet-I”>
    In this dialogue, Peter Pan and Arale-chan engage in a friendly discussion about AI’s capabilities and limitations. They exchange views on how AI could potentially assist or replace human creative work and ultimately recognise the importance of AI and humans working together.
    </voice>
    </speak>

カテゴリー: ピーターパンとアラレちゃんのAIについての討, 文章読み上げ, 生成AIの利用法, 音声合成マークアップ言語(SSML) パーマリンク