speak

Implementation Reference

• src/index.ts:51-69 (handler)

  Handler function that takes text input, uses VOICEVOX API via helpers to generate speech audio, saves it to /tmp/voicevox.wav, plays it using afplay, and returns success response.

  async ({ text }) => {
    const resolvedSpeakerId = Number(process.env.SPEAKER_ID);
    if (!resolvedSpeakerId || isNaN(resolvedSpeakerId)) {
      throw new Error("speaker_idが指定されてないか、環境変数SPEAKER_IDが不正です");
    }
    const query = await createAudioQuery(text, resolvedSpeakerId);
    const buffer = await synthesizeVoice(query, resolvedSpeakerId);
    const filePath = "/tmp/voicevox.wav";
    saveAudioFile(buffer, filePath);
    playAudio(filePath);
    return {
      content: [
        {
          type: "text",
          text: "OK",
        }
      ]
    };
  }

• src/index.ts:50-50 (schema)

  Zod schema defining the input parameter 'text' as a required string.

  { text: z.string() },

• src/index.ts:49-70 (registration)

  MCP server.tool call registering the 'speak' tool with its schema and handler function.

  server.tool("speak",
    { text: z.string() },
    async ({ text }) => {
      const resolvedSpeakerId = Number(process.env.SPEAKER_ID);
      if (!resolvedSpeakerId || isNaN(resolvedSpeakerId)) {
        throw new Error("speaker_idが指定されてないか、環境変数SPEAKER_IDが不正です");
      }
      const query = await createAudioQuery(text, resolvedSpeakerId);
      const buffer = await synthesizeVoice(query, resolvedSpeakerId);
      const filePath = "/tmp/voicevox.wav";
      saveAudioFile(buffer, filePath);
      playAudio(filePath);
      return {
        content: [
          {
            type: "text",
            text: "OK",
          }
        ]
      };
    }
  )

• src/api.ts:10-15 (helper)

  Helper function to create an audio synthesis query using VOICEVOX API.

  export async function createAudioQuery(text: string, speakerId: number) {
    const res = await fetch(`${VOICEVOX_API_URL}/audio_query?speaker=${speakerId}&text=${encodeURIComponent(text)}`, {
      method: "POST",
    });
    return await res.json();
  }

• src/api.ts:18-30 (helper)

  Helper function to synthesize audio from query using VOICEVOX API, applying custom speed scale from env.

  export async function synthesizeVoice(query: any, speakerId: number) {
    const speedScale = process.env.SPEED_SCALE ? Number(process.env.SPEED_SCALE) : 1.2;
    query.speedScale = speedScale;
    const res = await fetch(`${VOICEVOX_API_URL}/synthesis?speaker=${speakerId}`, {
      method: "POST",
      headers: {
        "Content-Type": "application/json",
      },
      body: JSON.stringify(query),
    });
    const audioBlob = await res.blob();
    return Buffer.from(await audioBlob.arrayBuffer());
  }