Skip to content

Pipecat integration · 3 min read · November 16, 2026

Build Pipecat voice agents with PyAI Hear and Speak

PyAI is listed in Pipecat's STT and TTS references. Watch the interruption demo, follow the setup guide, and run the existing WebRTC voice-agent example.
On this page
  1. See an interruption, then run the example
  2. Start with one usable result
By The PyAI team · Developer experience

The short answer

PyAI is listed in Pipecat's community integrations. Use Hear for streaming transcription and Speak for speech output while keeping your Pipecat transport, LLM, tools, and turn policy. The package and examples are maintained by PyAI.

The published Pipecat STT reference and TTS reference describe the two services. Start with our Pipecat guide for project checks, installation, pipeline placement, and troubleshooting.

See an interruption, then run the example

Watch the existing 34-second recording and choose your starter. It uses prerecorded synthetic input, live Hear and Speak, scripted replies, and VAD-driven interruption. The first reply stops when the second speaker begins. A human confirmed intelligible playback and the interruption. This is evidence for that recorded run, not a human-microphone or latency benchmark.

For a conversation, use the complete WebRTC example. It includes transport, VAD, an LLM, Hear, Speak, and disconnect cleanup. Follow its pinned Python and Pipecat dependencies; the newer framework setup has separate non-audio qualification.

Start with one usable result

  1. 1. Choose the path

    Use the downloadable text-to-WAV starter for a first speech file, or the WebRTC example for a conversational agent.

  2. 2. Verify the project

    Keep credentials on the server. Read GET /v1/me to confirm the organization, project, scopes, and effective limits. PyAI, your LLM, and hosting have separate billing.

  3. 3. Verify the output

    Download and play the first file, or connect to the WebRTC runner and check a short conversation. Confirm transcripts, playback, and cleanup before expanding the workload.

Install the published adapter with `pip install pipecat-pyai==0.1.5`. The Pipecat starter page links the guide, upstream references, recording, and runnable source. If you prefer a general voice-app scaffold, see the existing create-pyai-app walkthrough.

Primary sources

We cite the statute or ruling so you can verify every claim. Status changes, confirm currency with counsel.

Frequently asked questions

Does PyAI replace my Pipecat agent?

No. Hear and Speak are speech services inside your Pipecat pipeline. Your application retains its transport, LLM, tools, and turn policy.

Which example should I run first?

The downloadable starter writes a WAV file using a PyAI key. The complete WebRTC example adds transport, VAD, an LLM, speech input and output, and disconnect cleanup.

Does the recording qualify every microphone and framework version?

No. It demonstrates one run using prerecorded synthetic input and scripted replies. Test your microphone, transport, framework version, and workload before deployment.

All product

Get an API key. Make a phone agent.

Sandbox keys never bill. Fund a live key when you go to production.