Parley - Realtime AI Voice Agent Template
Realtime AI voice agent template. Rebrand and deploy in minutes.
Item Description
PARLEY: A COMPLETE REALTIME AI VOICE AGENT, READY TO REBRAND AND DEPLOY
Parley is a working AI voice agent with everything already connected. A visitor opens your web page, presses Start call, and talks to your assistant out loud. It listens, thinks, and answers back in a natural voice, with no typing and no long awkward pauses.
Speech recognition, the AI model, the voice, turn detection, the demo page, the free-trial timer and the deployment pipeline are all included. Add your own API keys, change a few settings, and you have a voice product you can put in front of customers the same day.
WHAT YOUR VISITORS EXPERIENCE
- A clean demo page with a Start call button
- A live agent that greets them by voice
- Clear call states: connecting, agent speaking, mic on, and time left
- Mute and End call controls
- A friendly warning when free time is almost up, followed by a polite goodbye
WHAT IS IN THE PACKAGE
- The voice agent, built in Python with LiveKit Agents
- A demo web page and token server, ready to deploy on Vercel
- A GitHub Actions workflow that deploys the agent every time you push to main
- A Dockerfile for container deployment
- Example environment files for both the agent and the web app
- A full README with setup, deployment and troubleshooting steps
- A license file for commercial use
THE VOICE PIPELINE
- Speech-to-text: Deepgram Nova-3
- AI model: Google Gemini, with an automatic backup model if the main one fails during a call
- Text-to-speech: Deepgram Aura-2, with a choice of voices
- Voice activity detection: Silero
- Turn detection: LiveKit multilingual model, so the agent knows when you have finished speaking
- Noise enhancement: ai-coustics, for cleaner audio on noisy lines
- Silence recovery: if the model ever produces no speech, the agent recovers on its own instead of going quiet
REBRAND WITHOUT EDITING CODE
Make Parley your own product by setting a few environment variables:
- AGENT_NAME: the name of your assistant
- AGENT_INSTRUCTIONS: its full personality, rules and knowledge
- AGENT_GREETING: the first thing it says when someone joins
- STT_LANGUAGE: the language it listens in
- TTS_MODEL: the voice it speaks with
- GEMINI_MODEL: the AI model it uses
Turn it into a receptionist, customer support agent, language tutor, subject tutor, sales assistant, onboarding guide, booking assistant, or any role you can describe in a prompt.
FREE DEMO LIMIT THAT TURNS VISITORS INTO CUSTOMERS
Each call has a time limit, 5 minutes by default. The agent warns the caller before time runs out, says goodbye, and ends the call. Your web page receives time_warning and time_limit events, so you can show a sign-up button at exactly the right moment. Optional per-device limits using Supabase stop one person from using your demo all day. Set the limit to 0 to switch it off.
KEEP YOUR COSTS UNDER CONTROL
- Optional access code for the demo page
- Rate limiting on token creation
- Optional per-device usage cap
- A pre-launch checklist in the README, including how to set spend limits with your providers
USE YOUR OWN FRONTEND
The included demo page is optional. You can connect your own website or mobile app by creating a LiveKit token that dispatches the agent. The README explains exactly what your app needs to do.
QUICK START
Run the agent locally:
cd agent
uv sync
uv run python src/agent.py download-files
uv run python src/agent.py dev
Copy .env.example to .env.local and add your keys first. Full deployment steps for LiveKit Cloud and Vercel are in the README.
WHAT YOU NEED
- Python with uv, and Node.js
- LiveKit CLI
- Accounts at LiveKit Cloud, Google AI Studio and Deepgram
- A Vercel account for the demo page and a GitHub account for auto-deploy
- A Supabase account, optional, for per-device demo limits
- These services bill you directly for usage. Parley itself is a one-time purchase.
WHO IT IS FOR
Developers, freelancers, agencies and founders who want to ship a voice product or a client demo fast, without building real-time audio from scratch.
LICENSE
Commercial license. Use and modify it for your own products and client projects. Reselling or republishing the source code as a template or starter kit is not allowed.
Features
- Realtime speech-in, speech-out AI voice agent
- Deepgram Nova-3 speech-to-text and Aura-2 text-to-speech
- Google Gemini with automatic backup model fallback
- Natural turn detection with smooth interruption handling
- Noise enhancement for clean audio
- Silence recovery so the agent never goes quiet
- Rebrand with settings: name, personality, greeting, language and voice
- Ready-made demo web page and token server for Vercel
- One-push deploy to LiveKit Cloud with GitHub Actions
- Dockerfile included
- Built-in free-demo timer with warning and goodbye
- Time warning and time limit events for sign-up buttons
- Optional per-device usage cap with Supabase
- Optional demo access code and token rate limiting
- Works with your own web or mobile frontend
- Full README with setup, deploy and troubleshooting guide
- Commercial license for your own and client projects
All Reviews
Verified source code
Free support included
Download code immediately
Quality guarantee for your satisfaction
Support: info@sellanycode.com
All Questions
- Verified source code
- Join 35401+ buyers which trust us!
- Download code immediately
Satisfaction guarantee policy.
Information
| Category | Scripts & Code / Python |
| First Release | 26 October 2026 |
| Files included | .html, Javascript .js, .py |
| File Size | 1 MB |
| Frameworks | Python |