Facial AR Engine & Virtual Camera

The project centers on a real-time facial-tracking engine that feeds a virtual camera, letting users stream or record video while high-quality 3D animations and objects track their expressions flawlessly. The engine must detect facial landmarks with precision, drive 3D meshes or rigs smoothly, and maintain low-latency performance (around 60 fps at 1080p is the working target). Once active, the output should appear as a standard webcam in apps like Zoom, OBS, or Teams.

Any robust stack is acceptable—C++/OpenCV, MediaPipe, Unity, Unreal Engine, or a custom OpenGL/Vulkan pipeline—so long as everything runs locally without licensing roadblocks. (Windows OS is ideal preferably a single executable file.
if you're familiar with the likes on snapchat & tiktok for example. these use filters that do exactly or near close to what im looking for.

There’s no fixed deadline; polish and clean, well-documented code take priority over speed. We can structure milestones for a proof-of-concept tracker, virtual-camera integration, an effect-authoring workflow, and final optimisation.

Deliverables expected:
• Full source code with build instructions
• Installable virtual-camera driver/component /software
• One showcase 3D animated mask or prop demonstrating the pipeline
• Concise technical documentation and an authoring guide for additional effects