It eliminates the need for local installation or GPU hardware to perform real-time face swapping and gives users a way to record the transformed video locally while keeping raw data on the client.
Live Face Swap is a browser-based tool that replaces the live video feed from a webcam with a reference image using an AI model. The user selects a preset or uploads a reference photo, chooses an engine (LFS Prism or LFS Mirror), and grants camera and microphone permissions. When the Face Swap button is pressed, the client streams raw video frames to a cloud inference service; the service runs the selected model, generates transformed frames at a rate defined by the credit system, and returns them to the browser for display. The transformed stream can be recorded locally, replayed, and saved as MP4/WebM without leaving the client. Credits are deducted per second of generation (1 sec for Prism, 20 sec for Mirror) and are shown before the session starts. The interface also provides diagnostics such as FPS, packet loss and latency for the current session, and supports output to OBS or other capture tools via a browser source. All processing occurs server‑side, while recordings and audio remain on the user's device.
Have feedback for the maker?
Sign in to leave a review, report a bug, or suggest a feature.
Qing Shao
Working on Live Face Swap, a browser workspace for creative camera effects. Based in Singapore.
