Eleven Labs Text2Speech

This notebook shows how to interact with the ElevenLabs API to achieve text-to-speech capabilities.

First, you need to set up an ElevenLabs account. You can follow the instructions here.

%pip install --upgrade --quiet  elevenlabs langchain-community

import os

os.environ["ELEVEN_API_KEY"] = ""

Usage

from langchain_community.tools import ElevenLabsText2SpeechTool

text_to_speak = "Hello world! I am the real slim shady"

tts = ElevenLabsText2SpeechTool()
tts.name

API Reference:ElevenLabsText2SpeechTool

'eleven_labs_text2speech'

We can generate audio, save it to the temporary file and then play it.

speech_file = tts.run(text_to_speak)
tts.play(speech_file)

Or stream audio directly.

tts.stream_speech(text_to_speak)

Use within an Agent

from langchain.agents import load_tools
from langchain_openai import OpenAI
from langgraph.prebuilt import create_react_agent

API Reference:load_tools | OpenAI | create_react_agent

llm = OpenAI(temperature=0)
tools = load_tools(["eleven_labs_text2speech"])
agent = create_react_agent(model=llm, tools=tools, debug=True)

audio_file = agent.run("Tell me a joke and read it out for me.")

> Entering new AgentExecutor chain...
Action:
\`\`\`
{
  "action": "eleven_labs_text2speech",
  "action_input": {
    "query": "Why did the chicken cross the playground? To get to the other slide!"
  }
}
\`\`\`


Observation: /tmp/tmpsfg783f1.wav
Thought: I have the audio file ready to be sent to the human
Action:
\`\`\`
{
  "action": "Final Answer",
  "action_input": "/tmp/tmpsfg783f1.wav"
}
\`\`\`



> Finished chain.

tts.play(audio_file)

Related

• Tool conceptual guide
• Tool how-to guides