How to Build Digital Humans From Priors to Photorealistic Avatars

Jul 16, 2026

Upload by Violet

This state-of-the-art report surveys controllable 3D human avatar creation, focusing on prior learning and avatar creation stages within a typical three-stage pipeline: learning priors of human appearance and motion, creating a personalized avatar, and animating it. It defines common avatar representations and builds a practical taxonomy organized along multiple axes, including body regions and the priors employed. It reviews full-body and head avatars as well as layered decompositions (hands, hair, garments), then summarizes key principles, core literature, and open research challenges toward future directions.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages49

Reading Time2 h 3 min

Introduction

Scope

Selection scheme

Structure of the report

Abstract

Overview of avatar creation stages

Prior learning and avatar creation focus

Representation taxonomy

Survey topics

Full-body and head avatars

Layered component representations

Principles and future directions

EUROGRAPHICS 2026 A. Bönsch and N. Umetani (Guest Editors) COMPUTER GRAPHICS forum Volume 45 (2026), Number 2 STAR – State of TheArt Report How to Build Digital Humans? From Priors to Photorealistic Avatars arXiv :2607 .0434 1v 1 [ cs .GR] 5 Jul 2026 Wojciech Zielonka 1†, Tobias Kirschstein2†, Timo Bolkart3†, Simon Giebenhain2, Vanessa Sklyarova5,8, Xiang Deng4 , Donglai Xiang6, Shunsuke Saito 1, Yebin Liu4 , Matthias Niessner2, Justus Thies7† 1Meta 2Technical University of Munich 3 Google 4Tsinghua University 5Max Planck Institute for Intelligent Systems 6NVIDIA 7Technical University of Darmstadt 8ETH Zurich † These authors contributed equally. [https://zielon.github.io/how-to-build-digital-humans/](https://zielon.github.io/how-to-build-digital-humans/) Figure 1: Building digital avatars requires considering many components, such as the human head, hair, hands, and garments, and ultimately full-body avatars. Much like da Vinci’s Vitruvian Man symbolizes the ideal proportions and unity of the human form, the construction of digital avatars demands a holistic approach in which each part integrates into a coherent whole. Each of these elements presents distinct challenges and numerous potential design choices across different stages. In this report, we introduce and examine the details of each creation stage, develop a practical taxonomy of the current literature, and discuss future directions in the field of digital avatars. The background image was generated using Google Gemini. [QKS∗24, SSS∗24, CHW∗25, GLK∗25, RWF∗25, SZH∗23, LZQ∗22, FPK∗24, BXX∗25, GBH23]. Abstract This state-of-the-art report provides an overview of controllable 3D human avatar creation. We describe current 3D avatar systems, which typically consist of three stages: (i) learning priors of human appearance and motion,(ii) creating a personal ized avatar, and (iii) animating the avatar. To limit the scope, we focus on the prior learning and avatar creation stages. We define current avatar representations and introduce a taxonomy that categorizes existing work along multiple axes, including body regions and employed priors. We review methods for full-body and head avatars, as well as layered representations that decompose the body into components such as hands, hair, and garments. Finally, we outline common underlying principles, reference key literature for newcomers, and discuss open challenges and future research directions. 1. Introduction For decades, the vision of constructing a photo-realistic digital representation of a person has driven research at the intersection of Computer Graphics, Computer Vision, and Machine Learning. Richard Feynman famously stated,“What I cannot create, I do not understand,” highlighting the epistemic value of creation as a means of understanding human appearance, behavior, and motion. But why do digital humans matter? There are already several industries that depend on digital humans, especially in the entertainment field (gaming, VFX), but also in the e-commerce field (virtual 2 of 49 W. Zielonka, T. Kirschstein, T. Bolkart, X. Deng, V. Sklyarova, S. Giebenhain, D. Xiang, S. Saito, Y. Liu, M. Niessner, J. Thies / STAR try-on) . As interactions shift into virtual and hybrid environments (VR/AR, telepresence, online collaboration), people need humancentered representations that preserve identity, expression, and social nuance. Digital humans become the bridge between physical and virtual communication, not only between humans, but also between humans and machines. A central question that is asked by many researchers, given the recent developments, is: Do we need 3D avatars in a world which is increasingly dominated by 2D representations? With the rise of image and video diffusion models, animatable portraits [XCG∗24, HGZ∗23], and stylized video avatars [OT24], 2D avatars have demonstrated their value. However, 3D avatars come with the promise of spatial interpretability and the ability to

FAQs

What three stages does the report describe in controllable 3D avatar creation?

The report outlines (i) learning priors of human appearance and motion, (ii) creating a personalized avatar, and (iii) animating the avatar. It concentrates on the prior learning and avatar creation stages.

What is the purpose of the report’s taxonomy?

It provides a practical taxonomy to categorize existing work along multiple axes. The axes include body regions and the priors used, helping new researchers navigate the field.

Which avatar types and body components are emphasized in the review?

The review covers full-body and head avatars, and also layered representations that decompose the body into components such as hands, hair, and garments.

Jul 16, 2026

Upload by Violet