菜单
Generate speech, transcribe audio, create voice agents, compose music, and manage voices using ElevenLabs MCP Server. Use when working with text-to-speech, speech-to-text, voice cloning, conversational AI agents, or music composition.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/pmarashian/cursor-agent-skills --skill elevenlabs-mcp命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
Setup and patterns for a Vite + React frontend with Capacitor and Ionic React, plus a separate Next.js backend in a monorepo. Use when creating or modifying a Capacitor mobile app with a Vite frontend and Next.js API, or when adding Capacitor to an existing Vite app with a Next.js backend.
Modify Cursor/VSCode user settings in settings.json. Use when the user wants to change editor settings, preferences, configuration, themes, font size, tab size, format on save, auto save, keybindings, or any settings.json values.
Automates browser interactions for web testing, form filling, screenshots, and data extraction. Use when the user needs to navigate websites, interact with web pages, fill forms, take screenshots, test web applications, or extract information from web pages.
基于 SOC 职业分类
正在显示 SKILL.md
| name | elevenlabs-mcp |
| description | Generate speech, transcribe audio, create voice agents, compose music, and manage voices using ElevenLabs MCP Server. Use when working with text-to-speech, speech-to-text, voice cloning, conversational AI agents, or music composition. |
| tags | ["audio","text-to-speech","speech-to-text","voice-cloning","ai-agents","music-composition","audio-processing"] |
Generate speech, transcribe audio, create voice agents, compose music, and manage voices using the ElevenLabs MCP Server.
The ElevenLabs MCP Server provides comprehensive tools for audio generation, transcription, voice management, conversational AI agents, and music composition. Many operations incur API costs - only use when explicitly requested by the user.
Server: ElevenLabs MCP (configured via MCP settings)
The MCP server is already configured and enabled. You can call these tools directly - they will be available in your tool list.
MANY TOOLS INCUR API COSTS - Only use when explicitly requested by the user:
Always check with the user before using cost-incurring tools unless they explicitly request them.
text_to_speech - Convert text to speech with voice selectiontext_to_voice - Create voice previews from text prompts (generates 3 variations)create_voice_from_preview - Add generated voice to libraryspeech_to_text - Transcribe audio files to textspeech_to_speech - Transform audio from one voice to anothertext_to_sound_effects - Generate sound effects from descriptionssearch_voices - Search your voice librarysearch_voice_library - Search entire ElevenLabs voice libraryget_voice - Get details of a specific voicevoice_clone - Create instant voice clone from audio filescreate_agent - Create a conversational AI agentadd_knowledge_base_to_agent - Add knowledge base to agentlist_agents - List all your agentsget_agent - Get agent detailsget_conversation - Get conversation transcriptlist_conversations - List agent conversationscompose_music - Generate music from prompt or composition plancreate_composition_plan - Create structured composition planisolate_audio - Isolate audio from a fileplay_audio - Play audio files (WAV, MP3)list_phone_numbers - List phone numbersmake_outbound_call - Make outbound call using agentlist_models - List available TTS modelscheck_subscription - Check subscription status and usage# Simple text-to-speech
result = text_to_speech(
text="Hello, this is a test of text-to-speech.",
voice_name="Adam", # or use voice_id
output_directory="./output"
)
# Returns file path to generated audio
You can use either voice_name or voice_id:
# Using voice name
text_to_speech(text="Hello", voice_name="Adam")
# Using voice ID
text_to_speech(text="Hello", voice_id="cgSgspJ2msm6clMCkdW9")
# Search for voices first
voices = search_voices(search="male professional")
voice_id = voices[0]["voice_id"]
text_to_speech(text="Hello", voice_id=voice_id)
Available models:
eleven_multilingual_v2 - High quality multilingual (29 languages)eleven_flash_v2_5 - Fastest, ultra-low latency (32 languages)eleven_turbo_v2_5 - Balanced quality and speed (32 languages)eleven_flash_v2 - Fast English-onlyeleven_turbo_v2 - Balanced English-onlyeleven_monolingual_v1 - Legacy Englishtext_to_speech(
text="Hello",
model_id="eleven_multilingual_v2",
language="en"
)
Key parameters for fine-tuning:
stability (0-1): Voice stability vs. emotional range (default: 0.5)similarity_boost (0-1): Adherence to original voice (default: 0.75)style (0-1): Style exaggeration (default: 0)use_speaker_boost (bool): Boost similarity to original speaker (default: true)speed (0.7-1.2): Speech speed (default: 1.0)output_format: Audio format (default: "mp3_44100_128")# Transcribe audio file
result = speech_to_text(
input_file_path="./audio/recording.mp3",
save_transcript_to_file=True,
output_directory="./transcripts"
)
# Returns transcript text and saves to file
# Transcribe with speaker identification
result = speech_to_text(
input_file_path="./audio/meeting.mp3",
diarize=True, # Identify different speakers
return_transcript_to_client_directly=True
)
# Returns transcript with speaker annotations
# Automatic language detection (default)
speech_to_text(input_file_path="./audio.mp3")
# Specify language
speech_to_text(
input_file_path="./audio.mp3",
language_code="es" # ISO 639-3 code
)
# Clone voice from audio samples
result = voice_clone(
name="My Custom Voice",
files=[
"./samples/sample1.mp3",
"./samples/sample2.mp3",
"./samples/sample3.mp3"
],
description="Professional male voice"
)
# Returns voice_id for use in TTS
# Generate voice previews
previews = text_to_voice(
voice_description="A warm, friendly female voice with a slight British accent",
text="Hello, this is a preview of the generated voice."
)
# Returns 3 preview variations with generated_voice_id
# Add to library
create_voice_from_preview(
generated_voice_id=previews["generated_voice_id"],
voice_name="Friendly British Voice",
voice_description="Warm, friendly female voice with British accent"
)
# Create conversational AI agent
agent = create_agent(
name="Customer Support Agent",
first_message="Hi, how can I help you today?",
system_prompt="You are a helpful customer support agent...",
voice_id="cgSgspJ2msm6clMCkdW9",
language="en"
)
# Returns agent_id
# Create agent first
agent_id = create_agent(...)["agent_id"]
# Add knowledge base
add_knowledge_base_to_agent(
agent_id=agent_id,
knowledge_base_name="Product Documentation",
input_file_path="./docs/product_manual.pdf"
)
# Or add from URL
add_knowledge_base_to_agent(
agent_id=agent_id,
knowledge_base_name="Company Website",
url="https://example.com/docs"
)
# Or add from text
add_knowledge_base_to_agent(
agent_id=agent_id,
knowledge_base_name="FAQ",
text="Q: What is your return policy? A: 30 days..."
)
# List available phone numbers
phone_numbers = list_phone_numbers()
# Make outbound call
make_outbound_call(
agent_id=agent_id,
agent_phone_number_id=phone_numbers[0]["id"],
to_number="+1234567890" # E.164 format
)
# List conversations
conversations = list_conversations(
agent_id=agent_id,
page_size=30
)
# Get conversation transcript
transcript = get_conversation(
conversation_id=conversations[0]["conversation_id"]
)
# Generate music from prompt
music = compose_music(
prompt="Upbeat electronic dance music with synthesizers",
music_length_ms=60000, # 60 seconds
output_directory="./music"
)
# Returns file path to generated music
# Create composition plan first
plan = create_composition_plan(
prompt="Epic orchestral piece with multiple movements",
music_length_ms=180000 # 3 minutes
)
# Generate music from plan
music = compose_music(
composition_plan=plan,
output_directory="./music"
)
# Isolate audio from file (remove background noise)
isolated = isolate_audio(
input_file_path="./audio/with_background.mp3",
output_directory="./audio/isolated"
)
# Transform audio to different voice
transformed = speech_to_speech(
input_file_path="./audio/original.mp3",
voice_name="Adam",
output_directory="./audio/transformed"
)
# Play audio file (WAV or MP3)
play_audio(input_file_path="./audio/sample.mp3")
# Generate sound effect
sound_effect = text_to_sound_effects(
text="Door creaking open slowly",
duration_seconds=2.0,
loop=False,
output_directory="./sounds"
)
text (required): Text to convertvoice_name or voice_id: Voice selectionmodel_id: TTS model (default: eleven_multilingual_v2)stability (0-1): Voice stabilitysimilarity_boost (0-1): Voice similarityspeed (0.7-1.2): Speech speedoutput_format: Audio formatoutput_directory: Save location (default: $HOME/Desktop)input_file_path (required): Audio file to transcribediarize (bool): Enable speaker diarizationlanguage_code: ISO 639-3 language code (auto-detect if omitted)save_transcript_to_file (bool): Save transcript to filereturn_transcript_to_client_directly (bool): Return text directlyname (required): Agent namefirst_message (required): First message agent sayssystem_prompt (required): System promptvoice_id: Voice for agent (default: cgSgspJ2msm6clMCkdW9)language: ISO 639-1 code (default: "en")llm: LLM backend (default: "gemini-2.0-flash-001")temperature (0-1): Response randomnessmax_duration_seconds: Max conversation length (default: 300)prompt or composition_plan (required): Music description or planmusic_length_ms: Duration in millisecondsoutput_directory: Save location (default: $HOME/Desktop)check_subscription() to monitor usageFile Output Behavior:
$HOME/Desktop regardless of output_directory parameteroutput_directory parameter is ignored{description}_{timestamp}.mp3Workflow:
$HOME/Desktop for generated filemv ~/Desktop/*.mp3 assets/audio/mv assets/audio/temp.mp3 assets/audio/final-name.mp3save_transcript_to_file=True for STT to keep transcriptsreturn_transcript_to_client_directly=True to get text immediatelysearch_voice_library() to browse all available voicesmax_duration_seconds to control costsmp3_22050_32 - Low quality MP3mp3_44100_32 - Standard quality MP3mp3_44100_64 - Good quality MP3mp3_44100_96 - High quality MP3mp3_44100_128 - Very high quality MP3 (default)mp3_44100_192 - Highest quality MP3 (Creator+ tier)pcm_8000 - PCM formatpcm_16000 - PCM formatpcm_22050 - PCM formatpcm_24000 - PCM formatpcm_44100 - PCM format (Pro+ tier)ulaw_8000 - μ-law format (Twilio compatible)alaw_8000 - A-law formatopus_48000_32 - Opus formatopus_48000_64 - Opus formatopus_48000_96 - Opus formatopus_48000_128 - Opus formatopus_48000_192 - Opus format# 1. Search for voice
voices = search_voices(search="professional female")
voice_id = voices[0]["voice_id"]
# 2. Generate speech
audio_file = text_to_speech(
text="Welcome to our application. How can I assist you?",
voice_id=voice_id,
model_id="eleven_turbo_v2_5",
stability=0.5,
similarity_boost=0.75,
speed=1.0,
output_directory="./assets/audio"
)
# 3. Use in application
# Reference: audio_file path
# 1. Create agent
agent = create_agent(
name="Product Support",
first_message="Hello! I'm here to help with product questions.",
system_prompt="You are a helpful product support agent...",
voice_id="cgSgspJ2msm6clMCkdW9"
)
agent_id = agent["agent_id"]
# 2. Add knowledge base
add_knowledge_base_to_agent(
agent_id=agent_id,
knowledge_base_name="Product Docs",
input_file_path="./docs/product.pdf"
)
# 3. List conversations later
conversations = list_conversations(agent_id=agent_id)
# 1. Create composition plan
plan = create_composition_plan(
prompt="Epic battle music with orchestral and electronic elements",
music_length_ms=120000 # 2 minutes
)
# 2. Generate music
music = compose_music(
composition_plan=plan,
output_directory="./assets/music"
)
# 3. Use in game/application
# Reference: music file path
Different features require different subscription tiers:
Use check_subscription() to see your current tier and usage limits.