• AUDIO
  • SPEED
  • CONTROL
  • HUMAN

How to stream real time speech with your AI

Connect Cartesia to your AI client to enable live audio. Your agent generates the text, and the audio flows immediately through a continuous stream.

Ask AI about this page

Short answer

How do I get my AI to talk in real time?

You use the tts_sse tool to pipe text directly into an audio stream via server-sent events. This allows your agent to start speaking before it even finishes writing a sentence. You'll get immediate, fluid audio output.

The outcomes

What you get from a live connection

Once you connect, your agent moves from text-only responses to active voice output.

AUDIO

Live voice output

The tts_sse tool converts text into a continuous stream of audio data. You get sound immediately instead of waiting for a full file to download.

SPEED

Low latency response

Because it uses server-sent events, the audio starts playing as the text is generated. This makes the conversation feel much more natural.

CONTROL

Voice personality

Your agent dictates the content, while the connector handles the audio synthesis. You control the tone through your prompts.

HUMAN

Manual control

The AI handles the speech generation, but you decide when to interrupt or change the topic. The connection stays open for your input.

The workflow

What your AI does when the text arrives.

The AI handles the conversation logic, then hands the words off to the connector for playback.

  1. Generate text

    Your agent processes your request and writes a response in plain text.

    tts_sse

  2. Stream audio

    The text is sent to the connector, which converts it into a live audio stream via server-sent events.

    tts_sse

  3. Play sound

    The audio data flows to your client, allowing the voice to start playing immediately.

    tts_sse

  4. Maintain flow

    The connection stays active so the AI can continue speaking as new text is produced.

    tts_sse

Try it

Copy these to start.

Use these prompts to test the real time audio streaming.

Starting points

These are opening lines to trigger the voice. Swap out the specific personas or topics to change how the AI sounds.

Cartesia Connector

You're all set. Choose your MCP client and follow the setup instructions.

Connector linkhttps://edge.vinkius.com/vk_preview_fqUkC4o6TEWq9kAU8DlY8MAb7pfJxo3R952S85p7/mcp

Claude Desktop

Follow the steps below to connect in seconds.

  1. 1In Claude Desktop, open Settings → Connectors.
  2. 2Click “Add custom connector” and paste the connector link above as the remote MCP server URL.
  3. 3Click Add and start a new chat — Cartesia capabilities are ready to use.
Configuration · claude_desktop_config.jsonCopy
{
  "mcpServers": {
    "cartesia-voice-ai-mcp": {
      "url": "https://edge.vinkius.com/vk_preview_fqUkC4o6TEWq9kAU8DlY8MAb7pfJxo3R952S85p7/mcp"
    }
  }
}
Copy into chat04

  • Read the latest updates from my project notes out loud.

  • Act as a helpful concierge and greet me using the live voice stream.

  • Summarize the contents of my recent email thread using real time speech.

  • Tell me a short story about a robot, and use the voice stream to narrate it.

  • Claude
  • ChatGPT
  • Cursor
  • VS Code
  • Windsurf
  • Claude Code
  • JetBrains
  • Cline

Start here

Connect Cartesia once, then ask.

Click the link to connect your account. Your credentials stay encrypted, and you can start using the voice tools in your preferred AI client immediately.

Connect Cartesia to your AI

FAQ

How the voice streaming behaves

  • 01

    Does the AI generate the audio files itself?

    No. The AI client generates the text, and the tts_sse tool handles the conversion to audio.

  • 02

    Can the AI change its own voice settings?

    No. The voice characteristics are determined by the connector settings and your initial prompts.

  • 03

    Is there a delay between text generation and speech?

    The use of server-sent events is designed to minimize this delay, allowing audio to start as text is being produced.

  • 04

    Can the AI delete my existing audio files?

    No. This connector only handles the generation and streaming of new audio.

  • 05

    Does this work with any AI client?

    It works with any MCP-compatible client like Claude, Cursor, or Windsurf.

  • More questions about Cartesia? The Connector page answers them. See everything the Cartesia Connector can do

Connect Cartesia to Claude, Cursor, ChatGPT & more