Official

Pipecat pipelines with Cartesia speech and transcription

Pipecat ships Cartesia text to speech and speech to text as first-party services that you install, add to your pipeline, and point at your API key.

Get an API keyRead the docs
install.sh
pip install "pipecat-ai[cartesia]"

bot.py
from pipecat.services.cartesia.tts import CartesiaTTSService
from pipecat.services.cartesia.turns.stt import CartesiaTurnsSTTService

stt = CartesiaTurnsSTTService(
    api_key=os.environ.get("CARTESIA_API_KEY"),
)

tts = CartesiaTTSService(
    api_key=os.environ.get("CARTESIA_API_KEY"),
)

pipeline = Pipeline(
    [
        transport.input(),
        stt,
        user_aggregator,
        llm,
        tts,
        transport.output(),
        assistant_aggregator,
    ]
)

Cartesia docs: Pipecat (read October 2026)

What you can do with Pipecat

What teams build when they put Cartesia and Pipecat together.

Two services, one extra

Installing pipecat-ai[cartesia] gives you both CartesiaTTSService and CartesiaTurnsSTTService from one package extra.

Turn detection from the transcriber

CartesiaTurnsSTTService is recommended over the older CartesiaSTTService for turn detection, and needs pipecat-ai[cartesia] 1.3.0+.

Start the reply earlier

The eager end-of-turn event starts the response slightly early, which Cartesia's guide says can cut about half a second.

Set it up

Official: the Cartesia text to speech and speech to text services are part of the Pipecat repository and install with one pip extra. Vendor details read in October 2026.

  1. Install

    Install pipecat-ai with the cartesia extra, plus the transport extra your example needs.

  2. Set CARTESIA_API_KEY

    Create a key in the Playground and export it.

  3. Add the services

    Put the speech to text service before the user aggregator and the text to speech service after the LLM.

  4. Run the example

    Pipecat's voice-cartesia-turns example is a working agent.

Documentation

Before you build

Pick a voice

Pass a voice ID from Cartesia's voice library when you create CartesiaTTSService in your pipeline.

Hosting

Run the pipeline on your own servers or on Pipecat Cloud. See the Daily page for the WebRTC transport.

More integrations

All integrations
Built in

Vercel AI SDK

Call Cartesia text to speech and speech to text from TypeScript through the AI SDK's Cartesia provider.

  • Sonic
  • Ink
See the integration
Official

LiveKit

Add Cartesia text to speech and speech to text to a LiveKit agent with the official plugin, or through LiveKit Inference without a separate Cartesia account.

  • Sonic
  • Ink
See the integration
Built in

Vapi

Pick a Cartesia voice or transcriber for a Vapi assistant, or connect your Cartesia API key to use the voices you've cloned.

  • Sonic
  • Ink
See the integration

Pipecat integration FAQs

How do I add Cartesia TTS to a Pipecat pipeline?

Run pip install "pipecat-ai[cartesia]", export CARTESIA_API_KEY, and create a CartesiaTTSService. Put it after the LLM and before transport.output() in the pipeline. The same extra installs CartesiaTurnsSTTService, which goes after transport.input() and before the user aggregator.

How do I set the voice and model in CartesiaTTSService?

Pass them in settings=CartesiaTTSService.Settings(voice="your-voice-id", model="sonic-3.6"). Pipecat's docs say a voice is required and the model defaults to sonic-3.6. The older voice_id and model constructor arguments are deprecated since 0.0.105 and will be removed in 2.0.0, so move them into Settings when you upgrade.

Which Cartesia STT service should I use in Pipecat?

CartesiaTurnsSTTService. It runs Ink 2, detects the end of each turn on Cartesia's side, and Cartesia's guide recommends it over the older CartesiaSTTService. It needs pipecat-ai[cartesia] 1.3.0 or later, and Pipecat's voice-cartesia-turns example uses it.

Should I use the WebSocket or HTTP Cartesia service in Pipecat?

Use CartesiaTTSService, the WebSocket service, for live voice agents. Pipecat's docs recommend it for interactive use, and it streams audio and returns word timestamps. CartesiaHttpTTSService returns each request's audio in one piece, with no word timestamps.

How much does Cartesia cost in a Pipecat pipeline?

Pipecat calls Cartesia's API directly, so you pay API credits on your plan, not the voice agent rate. Sonic text to speech costs 1 credit per character and Ink speech to text 3 credits per second of audio. The Free plan includes 20K credits a month, and paid plans start at $5/month. Your LLM and transport bill separately.

Is there a full Pipecat walkthrough?

Yes. The guide to building a voice agent with Pipecat and Cartesia runs the voice-cartesia-turns example in a browser, swaps the voice and LLM, adds eager turn ends, and puts the agent on a phone number.

Get started today

Talk to an expert.

Connect with a member of our team and learn how Cartesia can help you build world-class voice experiences.

Contact Sales

Start building.

Access our models via API and bring a voice agent into production in minutes.

Try Cartesia