suno_stems_music

Instructions

Splits the audio into separate tracks for vocals and instrumentals,
useful for remixing, karaoke, or isolating specific parts.

Use this when:
- You want to separate vocals from instrumentals
- You need individual stem tracks for mixing
- You want to create a karaoke version

Returns:
    Task ID and stem separation results with individual track URLs.

Implementation Reference

• tools/audio_tools.py:464-493 (handler)

  The main handler function for the suno_stems_music tool. Decorated with @mcp.tool(), it takes an audio_id (and optional callback_url), calls client.generate_audio with action='stems', and returns the formatted result.

  @mcp.tool()
  async def suno_stems_music(
      audio_id: Annotated[
          str,
          Field(description="ID of the audio to separate into stems."),
      ],
      callback_url: Annotated[
          str | None,
          Field(description="Webhook callback URL for asynchronous notifications."),
      ] = None,
  ) -> str:
      """Separate a song into individual stems (vocals and instruments).
  
      Splits the audio into separate tracks for vocals and instrumentals,
      useful for remixing, karaoke, or isolating specific parts.
  
      Use this when:
      - You want to separate vocals from instrumentals
      - You need individual stem tracks for mixing
      - You want to create a karaoke version
  
      Returns:
          Task ID and stem separation results with individual track URLs.
      """
      result = await client.generate_audio(
          action="stems",
          audio_id=audio_id,
          callback_url=callback_url,
      )
      return format_audio_result(result)

• main.py:210-213 (registration)

  Registration of suno_stems_music in the server card tool list (main.py HTTP route), listing the tool with a description for client discovery.

  {
      "name": "suno_stems_music",
      "description": "Separate into vocal and instrument stems",
  },

• tools/audio_tools.py:464-493 (registration)

  The @mcp.tool() decorator registers suno_stems_music as an MCP tool with the FastMCP server instance.

  @mcp.tool()
  async def suno_stems_music(
      audio_id: Annotated[
          str,
          Field(description="ID of the audio to separate into stems."),
      ],
      callback_url: Annotated[
          str | None,
          Field(description="Webhook callback URL for asynchronous notifications."),
      ] = None,
  ) -> str:
      """Separate a song into individual stems (vocals and instruments).
  
      Splits the audio into separate tracks for vocals and instrumentals,
      useful for remixing, karaoke, or isolating specific parts.
  
      Use this when:
      - You want to separate vocals from instrumentals
      - You need individual stem tracks for mixing
      - You want to create a karaoke version
  
      Returns:
          Task ID and stem separation results with individual track URLs.
      """
      result = await client.generate_audio(
          action="stems",
          audio_id=audio_id,
          callback_url=callback_url,
      )
      return format_audio_result(result)

• core/client.py:172-176 (helper)

  The generate_audio method on SunoClient that suno_stems_music calls internally. It sends a POST request to /suno/audios with the provided params (including action='stems').

  # Convenience methods for specific endpoints
  async def generate_audio(self, **kwargs: Any) -> dict[str, Any]:
      """Generate audio using the audios endpoint."""
      logger.info(f"🎵 Generating audio with action: {kwargs.get('action', 'generate')}")
      return await self.request("/suno/audios", self._with_async_callback(kwargs))

• core/utils.py:77-90 (helper)

  The format_audio_result utility used by suno_stems_music to format the API response as JSON with async submission guidance.

  def format_audio_result(data: dict[str, Any]) -> str:
      """Format audio generation result as JSON.
  
      Args:
          data: API response dictionary
  
      Returns:
          JSON string representation of the result
      """
      return json.dumps(
          _with_submission_guidance(data, "suno_get_task", "suno_get_tasks_batch"),
          ensure_ascii=False,
          indent=2,
      )