Is your feature request related to a problem? Please describe.
There is currently no end-to-end sample in this repo that demonstrates how to run a Copilot Studio agent with real-time voice using the Azure Voice Live API.
For makers/devs it’s hard to get started because the required wiring (streaming audio, turn-taking, barge-in, auth, configuration, and deployment) is not obvious without a reference implementation.
Describe the solution you'd like
Please add a working sample/template that shows a Copilot Studio agent running with Azure Voice Live API, including:
- A minimal web client (browser microphone) for real-time conversation (push-to-talk or open-mic).
- A lightweight backend (Azure Functions / Node / .NET) that connects to Azure Voice Live API for low-latency audio streaming and forwards text/turns to the Copilot Studio agent.
- Clear setup steps + required environment variables (keys, endpoints, region, agent/channel configuration).
- Support for:
- streaming input (mic) + streaming output (TTS)
- barge-in (interrupting TTS when user starts speaking)
- basic turn-taking and session lifecycle (start/stop)
- logging/troubleshooting guidance (common errors and how to debug)
- Optional but very helpful:
- IaC (Bicep/Terraform) to deploy required Azure resources
- a simple architecture diagram and sequence diagram
Describe alternatives you've considered
- Using Speech SDK samples or custom integrations, but they don’t provide a Copilot Studio-specific, copy/paste starter template in this repo.
- Using non-real-time voice approaches (record-then-send) which increases latency and reduces user experience.
Additional context
Use cases:
- Voice-enabled Copilot Studio agent in a web app (no phone/PSTN required).
- A reference implementation would help teams standardize on a supported pattern for real-time voice (auth, session management, and reliable streaming).
If needed, I can help test the sample on a real tenant and provide feedback/PRs for documentation.
Is your feature request related to a problem? Please describe.
There is currently no end-to-end sample in this repo that demonstrates how to run a Copilot Studio agent with real-time voice using the Azure Voice Live API.
For makers/devs it’s hard to get started because the required wiring (streaming audio, turn-taking, barge-in, auth, configuration, and deployment) is not obvious without a reference implementation.
Describe the solution you'd like
Please add a working sample/template that shows a Copilot Studio agent running with Azure Voice Live API, including:
Describe alternatives you've considered
Additional context
Use cases:
If needed, I can help test the sample on a real tenant and provide feedback/PRs for documentation.