Wojciech ZielonkaPersonal website of Wojciech Zielonka, AI Research Scientist at Meta in Pittsburgh, PA, working on conversational agents, LLMs, multimodal AI, embodied AI, and digital humans — human motion, behavior, reasoning and appearance.https://wojciechzielonka.com/2026-10-08T00:12:56+00:00Wojciech Zielonkahttps://wojciechzielonka.comNo Distillation Needed: Single-Pass Real-Time Talking Heads via Acausal Noise Shapinghttps://wojciechzielonka.com/facegan/2026-10-07T10:00:00+00:002026-10-07T10:00:00+00:00FaceGAN (arXiv 2026) — a streaming single-pass GAN that maps speech to facial expression and head pose in real time, using latency-free acausal noise shaping for smooth, expressive motion.How to Build Digital Humans? From Priors to Photorealistic Avatarshttps://wojciechzielonka.com/digital-humans/2026-02-22T10:00:00+00:002026-02-22T10:00:00+00:00STAR report (Eurographics 2026) surveying how to build photorealistic digital humans — from priors and geometry to appearance and avatar synthesis.PhysHead: Simulation-Ready Gaussian Head Avatarshttps://wojciechzielonka.com/physhead/2026-02-21T10:00:00+00:002026-02-21T10:00:00+00:00PhysHead (CVPR 2026) — simulation-ready Gaussian head avatars with integrated physical dynamics for realistic motion and contact.SynShot: Synthetic Prior for Few-Shot Drivable Head Avatar Inversionhttps://wojciechzielonka.com/synshot/2025-01-12T10:32:20+00:002025-01-12T10:32:20+00:00SynShot (CVPR 2025) — synthetic prior enables few-shot drivable Gaussian head avatar inversion from minimal input images.GEM: Gaussian Eigen Models for Human Headshttps://wojciechzielonka.com/gem/2024-03-19T10:32:20+00:002024-03-19T10:32:20+00:00GEM (CVPR 2025) — compresses 3D Gaussian head avatars into a low-dimensional PCA-style eigen basis for efficient editing and animation.D3GA: Drivable 3D Gaussian Avatarshttps://wojciechzielonka.com/d3ga/2023-11-12T10:32:20+00:002023-11-12T10:32:20+00:00D3GA (3DV 2025) — drivable 3D Gaussian splatting avatars combined with cage-based deformations for full-body human representation.GAN-Avatar: Controllable Personalized GAN-based Human Head Avatarhttps://wojciechzielonka.com/ganavatar/2023-11-01T10:32:20+00:002023-11-01T10:32:20+00:00GAN-Avatar (3DV 2024) — controllable personalized GAN-based human head avatars built from in-the-wild monocular video.INSTA: Instant Volumetric Head Avatarshttps://wojciechzielonka.com/insta/2022-11-14T10:32:20+00:002022-11-14T10:32:20+00:00INSTA (CVPR 2023) — photorealistic volumetric head avatar reconstruction from a monocular video in under 10 minutes using neural radiance fields.MICA: Towards Metrical Reconstruction of Human Faceshttps://wojciechzielonka.com/mica/2022-07-09T10:32:20+00:002022-07-09T10:32:20+00:00MICA (ECCV 2022) — metrically accurate 3D face shape estimation from a single 2D image, leveraging face recognition features.Vulkan Path Tracer Sandboxhttps://wojciechzielonka.com/vulkan/2021-10-10T10:32:20+00:002021-10-10T10:32:20+00:00Vulkan-based real-time physically-based path tracer for shader prototyping with custom integrators, materials, and scene assets.Software Rasterizerhttps://wojciechzielonka.com/rasterizer/2021-09-01T10:32:20+00:002021-09-01T10:32:20+00:00From-scratch CPU software rasterizer in C++ using SIMD intrinsics for fast triangle rendering, depth testing, and texture sampling.Parallel Mesh Simplification For Planar Surfaceshttps://wojciechzielonka.com/mesh-simplification/2021-05-14T10:32:20+00:002021-05-14T10:32:20+00:00Parallel triangle mesh simplification using quadric error metric and adaptive thresholding to compress planar surfaces while preserving detail.Audio-driven Neural Rendering of Portrait Videoshttps://wojciechzielonka.com/face-neural-rendering/2021-05-05T10:32:20+00:002021-05-05T10:32:20+00:00Master thesis (TUM, 2021) — photorealistic audio-driven portrait video synthesis using neural rendering of 3DMM-based face geometry.RGB Face Tracking and Reconstructionhttps://wojciechzielonka.com/face2face/2020-02-08T10:32:20+00:002020-02-08T10:32:20+00:00Photometric markerless 3D face tracking using analysis-by-synthesis 3DMM optimization on a CUDA/OpenGL GPU pipeline.