Deploy Real-time Video-to-Video Model

Run a real-time video-to-video pipeline that returns YOLO detections on your stream.

This example turns a live webcam feed into annotated video using YOLO. WebRTC is the transport layer for low-latency, real-time streaming.

🚀 Try this Example

View the complete source code on GitHub.

Steps to run:

1. Install fal:

pip install fal

2. Authenticate:

fal auth login

3. Clone the demos repo and install dependencies:

git clone https://github.com/fal-ai-community/fal-demos.git
cd fal-demos
pip install -e .

4. Run the backend (local dev):

fal run fal_demos/video/yolo_webcam_webrtc/yolo.py::WebcamWebRtc

5. Run the frontend:

cd fal_demos/video/yolo_webcam_webrtc/frontend
npm install
FAL_KEY=myfalkey npm run dev

Open the Vite dev server in your browser and set the Endpoint field to the full real-time endpoint (for example: myuser/myapp/webrtc).

This deployment flow is the same as other serverless examples, like Deploy Text-to-Image Model, but the request path is a real-time endpoint.

Deploy to fal

To host the backend, deploy the app and use the resulting endpoint in the UI:

fal deploy fal_demos/video/yolo_webcam_webrtc/yolo.py::WebcamWebRtc

The real-time endpoint will be your-username/your-app/webrtc.

How it works

• The browser streams your webcam to the app in real-time (WebRTC transport).
• The backend runs YOLO on each frame and draws detection boxes.
• The annotated stream is sent back to the browser over the same connection.