Run your first Pipecat bot in under 5 minutes. This example creates a voice AI bot that you can talk to in your browser.
Pipecat requires Python 3.10 or newer. Check your version:
python --versionIf you need to upgrade Python, we recommend using a version manager like uv or pyenv.
Pipecat orchestrates different AI services in a pipeline, ensuring low latency communication. In this quickstart example, we'll use:
- Deepgram for Speech-to-Text transcriptions
- OpenAI for LLM inference
- Cartesia for Text-to-Speech audio generation
Have your API keys ready. We'll add them to your .env shortly.
-
Clone this repository
git clone https://github.com/pipecat-ai/pipecat-quickstart.git cd pipecat-quickstart -
Configure environment variables
Create a
.envfile:cp env.example .env
Then, add your API keys:
DEEPGRAM_API_KEY=your_deepgram_api_key OPENAI_API_KEY=your_openai_api_key CARTESIA_API_KEY=your_cartesia_api_key -
Set up a virtual environment and install dependencies
python -m venv .venv source .venv/bin/activate # On Windows: .venv\Scripts\activate pip install -r requirements.txt
Using
uv? Create your venv using:uv sync -
Run the example
python bot.py
Using
uv? Run your bot using:uv run bot.py. -
Connect and test
Open http://localhost:7860/client in your browser and click
Connectto start talking to your bot.💡 First run note: The initial startup may take ~15 seconds as Pipecat downloads required models, like the Silero VAD model.
- Browser permissions: Make sure to allow microphone access when prompted by your browser.
- Connection issues: If the WebRTC connection fails, first try a different browser. If that fails, make sure you don't have a VPN or firewall rules blocking traffic. WebRTC uses UDP to communicate.
- Audio issues: Check that your microphone and speakers are working and not muted.
- Read the docs: Check out Pipecat's docs for guides and reference information.
- Join Discord: Join Pipecat's Discord server to get help and learn about what others are building.