Large-Scale Codec Avatars: Surprising Results from Pretraining at Scale

Forum topic · 小凯 · 2026-04-05

Summary

A forum post introduces the paper 'Large-Scale Codec Avatars (LCA)' (arXiv:2604.02320), a high-fidelity, full-body 3D avatar model that generalizes to world-scale populations in a feedforward manner. The work addresses the classic trade-off between fidelity and generalization in 3D avatar modeling: multi-view studio data yields high fidelity but poor generalization, while models trained on millions of in-the-wild samples generalize widely but produce low-quality avatars. Inspired by large language models and vision foundation models, the authors introduce a pre/post-training paradigm for 3D avatars: pretraining on 1 million in-the-wild videos to learn broad appearance and geometry priors, followed by post-training on curated high-quality data to boost expressivity and fidelity. LCA generalizes across hairstyles, clothing, and demographics, supports fine-grained facial expressions and finger-level articulation, and preserves identity well. Emergent capabilities include relightability, loose garment handling, and zero-shot robustness to stylized imagery without direct supervision.

Paper Overview

Research area: CV/Graphics Authors: Junxuan Li, Rawal Khirodkar, Chengan He Published: 2026-04-02 arXiv: 2604.02320

Abstract

High-quality 3D avatar modeling faces a critical trade-off between fidelity and generalization. On the one hand, multi-view studio data enables high-fidelity modeling of humans with precise control over expressions and poses, but it struggles to generalize to real-world data due to limited scale and the domain gap between the studio environment and the real world. On the other hand, recent large-scale avatar models trained on millions of in-the-wild samples show promise for generalization across a wide range of identities, yet the resulting avatars are often of low-quality due to inherent 3D ambiguities.

To address this, the authors present Large-Scale Codec Avatars (LCA), a high-fidelity, full-body 3D avatar model that generalizes to world-scale populations in a feedforward manner, enabling efficient inference. Inspired by the success of large language models and vision foundation models, this work presents, for the first time, a pre/post-training paradigm for 3D avatar modeling at scale:

Pretraining on 1M in-the-wild videos to learn broad priors over appearance and geometry
Post-training on high-quality curated data to enhance expressivity and fidelity

LCA generalizes across hair styles, clothing, and demographics while providing precise, fine-grained facial expressions and finger-level articulation control, with strong identity preservation.

Emergent Capabilities

Notably, the authors observe generalization to unconstrained inputs that emerges without direct supervision:

Relightability
Support for loose garments
Zero-shot robustness to stylized imagery

---

*Auto-collected on 2026-04-05*