No cloud inference
Your face, camera, and call frames are processed on your own Windows machine.
Local identity system / Windows
A single image becomes a live camera persona you can use in Zoom, Google Meet, WhatsApp, and other video calls. Everything happens on your GPU.
Windows may show “Unknown publisher.” New here? Follow the first-call guide
01 / LOCAL PIPELINE
The physical camera stays behind Huantu’s final identity boundary. Calling apps see only transformed frames, a short approved hold, or black.
Your face, camera, and call frames are processed on your own Windows machine.
No Python environments, CUDA toolkits, OBS scenes, or manual camera plumbing.
If transformed output becomes unsafe, Huantu closes the route instead of exposing you.
03 / FAIL-CLOSED
Tracking loss, engine failure, restart, or panic never routes the physical camera through to the call. The boundary changes directly from transformed output to black.
04 / THE INSTALL MIRACLE
Huantu detects the machine, prepares the local runtime, installs its camera, and surfaces problems in plain language. You operate identities—not infrastructure.
05 / ONE CAMERA, EVERYWHERE
If an app can select a webcam, it can select Huantu Camera.