什么是 ElevenLabs 语音与转录？

ElevenLabs 语音技能是一个全面的音频处理工具，专为需要通过人类声音进行交流或理解人类声音的 AI 智能体设计。它利用 ElevenLabs API 提供两项主要服务：使用先进神经模型的高质量文本转语音 (TTS) 和通过 Scribe 引擎实现的语音转文本 (STT)。这一集成允许 Openclaw Skills 生态系统内的开发者创建高度互动、支持多语言且具备情感表达和精确音频理解能力的智能体。

通过使用此技能，智能体可以在 eleven_turbo_v2_5（追求速度）或 eleven_multilingual_v2（支持全球化）等各种模型之间切换。它对于需要自然旁白或处理来自 T@elegrimm 等平台语音备忘录的工作流特别有效。该技能将复杂的音频任务简化为直观的 CLI 命令和 Python 方法，使其成为现代 Openclaw Skills 实现的基础组件。

下载入口:https://github.com/openclaw/skills/tree/main/skills/jeffpignataro/miranda-elevenlabs-speech

安装与下载

1. ClawHub CLI

从源直接安装技能的最快方式。

2. 手动安装

将技能文件夹复制到以下位置之一

优先级：工作区 > 本地 > 内置

3. 提示词安装

将此提示词复制到 OpenClaw 即可自动安装。

ElevenLabs 语音与转录 应用场景

• 为客户支持机器人和个人助手创建自动语音回复。
• 转录用户的语音消息，以便大语言模型 (LLM) 处理语音优先的输入。
• 为国际观众生成 99 种不同语言的本地化语音内容。
• 使用自定义语音稳定性设置，实现故事讲述或新闻发布的生动旁白。

1. 用户提供用于合成的文本字符串或用于转录的音频文件（如 .mp3, .ogg）。
2. 对于合成，该技能调用 ElevenLabs TTS API，并带有 voice_id、stability 和 similarity_boost 等参数。
3. 对于转录，ElevenLabs Scribe 引擎处理音频文件，可选择使用语言提示或说话人识别。
4. 系统将生成的音频保存到指定路径，或将转录的文本返回给智能体工作流。
5. 智能体随后可以使用输出发送语音回复或继续进行基于文本的推理。

ElevenLabs 语音与转录 配置指南

要在您的 Openclaw Skills 环境中开始使用此技能，请配置您的 API 凭据：

export ELEVENLABS_API_KEY="your_api_key_here"

您也可以将其存储在工作区根目录的 .env 文件中。确保您的脚本目录中已包含 elevenlabs_speech.py 和 elevenlabs_scribe.py 以处理执行逻辑。

ElevenLabs 语音与转录 数据架构与分类体系

该技能使用以下结构管理音频数据和元数据：

支持的音频格式包括 mp3, mp4, wav 和 ogg（针对 T@elegrimm 进行了优化）。

name: elevenlabs-speech
description: Text-to-Speech and Speech-to-Text using ElevenLabs AI. Use when the user wants to convert text to speech, transcribe voice messages, or work with voice in multiple languages. Supports high-quality AI voices and accurate transcription.

ElevenLabs Speech

Complete voice solution — both TTS and STT using one API:

• TTS: Text-to-Speech (high-quality voices)
• STT: Speech-to-Text via Scribe (accurate transcription)

Quick Start

Environment Setup

Set your API key:

export ELEVENLABS_API_KEY="sk_..."

Or create .env file in workspace root.

Text-to-Speech (TTS)

Convert text to natural-sounding speech:

python scripts/elevenlabs_speech.py tts -t "Hello world" -o greeting.mp3

With custom voice:

python scripts/elevenlabs_speech.py tts -t "Hello" -v "voice_id_here" -o output.mp3

List Available Voices

python scripts/elevenlabs_speech.py voices

Using in Code

from scripts.elevenlabs_speech import ElevenLabsClient

client = ElevenLabsClient(api_key="sk_...")

# Basic TTS
result = client.text_to_speech(
    text="Hello from zerox",
    output_path="greeting.mp3"
)

# With custom settings
result = client.text_to_speech(
    text="Your text here",
    voice_id="21m00Tcm4TlvDq8ikWAM",  # Rachel
    stability=0.5,
    similarity_boost=0.75,
    output_path="output.mp3"
)

# Get available voices
voices = client.get_voices()
for voice in voices['voices']:
    print(f"{voice['name']}: {voice['voice_id']}")

Popular Voices

Voice Settings

• stability (0-1): Lower = more emotional, Higher = more stable
• similarity_boost (0-1): Higher = closer to original voice

Default: stability=0.5, similarity_boost=0.75

Models

• eleven_turbo_v2_5 - Fast, high quality (default)
• eleven_multilingual_v2 - Best for non-English
• eleven_monolingual_v1 - English only

Integration with T@elegrimm

When user sends text and wants voice reply:

# Generate speech
result = client.text_to_speech(text=user_text, output_path="reply.mp3")

# Send via T@elegrimm message tool with media path
message(action="send", media="path/to/reply.mp3", as_voice=True)

Pricing

Check https://elevenlabs.io/pricing for current rates. Free tier available!

Speech-to-Text (STT) with ElevenLabs Scribe

Transcribe voice messages using ElevenLabs Scribe:

Transcribe Audio

python scripts/elevenlabs_scribe.py voice_message.ogg

With specific language:

python scripts/elevenlabs_scribe.py voice_message.ogg --language ara

With speaker diarization (multiple speakers):

python scripts/elevenlabs_scribe.py voice_message.ogg --speakers 2

Using in Code

from scripts.elevenlabs_scribe import ElevenLabsScribe

client = ElevenLabsScribe(api_key="sk-...")

# Basic transcription
result = client.transcribe("voice_message.ogg")
print(result['text'])

# With language hint (improves accuracy)
result = client.transcribe("voice_message.ogg", language_code="ara")

# With speaker detection
result = client.transcribe("voice_message.ogg", num_speakers=2)

Supported Formats

• mp3, mp4, mpeg, mpga, m4a, wav, webm
• Max file size: 100 MB
• Works great with T@elegrimm voice messages (.ogg)

Language Support

Scribe supports 99 languages including:

• Arabic (ara)
• English (eng)
• Spanish (spa)
• French (fra)
• And many more...

Without language hint, it auto-detects.

Complete Workflow Example

User sends voice message → You reply with voice:

from scripts.elevenlabs_scribe import ElevenLabsScribe
from scripts.elevenlabs_speech import ElevenLabsClient

# 1. Transcribe user's voice message
stt = ElevenLabsScribe()
transcription = stt.transcribe("user_voice.ogg")
user_text = transcription['text']

# 2. Process/understand the text
# ... your logic here ...

# 3. Generate response text
response_text = "Your response here"

# 4. Convert to speech
tts = ElevenLabsClient()
tts.text_to_speech(response_text, output_path="reply.mp3")

# 5. Send voice reply
message(action="send", media="reply.mp3", as_voice=True)

Pricing

Check https://elevenlabs.io/pricing for current rates:

TTS (Text-to-Speech):

• Free tier: 10,000 characters/month
• Paid plans available

STT (Speech-to-Text) - Scribe:

• Free tier available
• Check website for current pricing