Real-Time AI Video / Virtual Camera Developer

We are looking for an experienced AI / Computer Vision developer to build a real-time video transformation system. ## Project Goal The application should: * Capture live video from a webcam. * Process the webcam feed using AI in real time. * Transform the person in the video into a predefined, authorized digital identity/avatar while preserving natural head movement, facial expressions, and motion. * Generate the transformed video with low latency. * Output the processed video as a virtual camera that can be selected by compatible video applications. * Maintain stable video quality during continuous use. The system is intended for authorized avatar / visual-effects use. It is not intended for identity verification, KYC, authentication, or impersonation. ## Technical Expectations We are open to the developer recommending the best architecture and technology stack. Relevant experience may include: * Python and/or C++ * PyTorch * OpenCV * Real-time generative AI * Face tracking / facial landmarks * Face or avatar reenactment * Video-to-video generation * CUDA / GPU inference optimization * OBS / virtual camera integration * ONNX / TensorRT optimization Existing open-source models and frameworks may be used where appropriate. We do not require the AI model to be developed from scratch. ## Performance Target Ideally: * Real-time or near-real-time processing * Target: approximately 25–30 FPS * Low latency * 720p minimum, preferably 1080p * Natural facial expressions and head movement * Stable output during longer sessions GPU acceleration is expected. ## Phase 1 — Proof of Concept For the first milestone, we want a working prototype demonstrating: Webcam Input → Real-Time AI Processing → Transformed Video → Virtual Camera Output The PoC does not need a polished user interface. The most important factors are image quality, stability, FPS, and latency. ## Future Phase If the PoC is successful, we may continue with: * Performance optimization * Simple desktop UI * Multiple authorized avatars * GPU selection/settings * Resolution and FPS settings * Easy avatar/profile management * Windows packaging/installer ## When Applying Please answer the following: 1. Have you previously developed real-time face/avatar/video transformation systems? 2. Can you share a relevant demo or portfolio example? 3. Which model/framework would you recommend for this project? 4. What GPU would you recommend for 1080p real-time processing? 5. What FPS and end-to-end latency do you realistically expect? 6. Would you use an existing model or develop/train a custom model? 7. What would you estimate for the PoC development time and cost? Please briefly describe the architecture you would use before providing your estimate.