SpatialAvatar-0: High-Quality 4D Head Avatar with Multi-Stage Reconstruction

2026-06-14

AI Summary

This work addresses key limitations in existing 4D avatar methods—namely domain bias, the absence of end-to-end shared representations, and time-consuming per-subject optimization that disrupts Gaussian layout. To overcome these challenges, we propose a multi-stage reconstruction framework that binds Gaussians to a FLAME mesh, integrating feedforward generation with lightweight per-subject refinement. Our approach achieves the first end-to-end unified two-stage pipeline through parameter-free K-source mean pooling, a two-phase training strategy, and a layout-preserving fine-tuning mechanism that effectively maintains Gaussian distribution structure. In cross-domain zero-shot evaluation, our method improves PSNR by 1.5 dB over GAGAvatar; on monocular benchmarks, it surpasses GeoAvatar by 1.3 dB while accelerating training by 60×.

Abstract

High-quality 4D head avatars from one or a few source portraits are central to telepresence, AR/VR, and digital-human interaction. 3D Gaussian Splatting (3DGS) has emerged as the dominant representation, with two complementary regimes (generalizable feed-forward predictors and per-subject refiners) maturing in parallel. However, existing feed-forward predictors are trained on a single dataset family with a hard-coded source count, inheriting the corresponding domain bias. Per-subject refiners require 300K--600K iterations and rely on adaptive densification that destroys upstream Gaussian layouts, preventing the two regimes from sharing a representation end-to-end. To bridge both regimes we propose SpatialAvatar-0 on a shared FLAME-mesh-bound Gaussian representation: a feed-forward generator with a parameter-free K-source mean-pool and a monocular-temporal to multi-view-spatial two-phase schedule that anchors against identity-prior collapse onto the smaller multi-view set. We further introduce a 10K-iter layout-preserving per-subject refinement loop that freezes the FLAME-binding and Gaussian count and replaces densification with a three-component anti-spike regularization. On VFHQ/HDTF cross-domain zero-shot we surpass the in-domain leader GAGAvatar by +1.5 dB PSNR despite never training on either test domain, and on the SplattingAvatar monocular benchmark we lead every reported metric, surpassing the 300K-iter GeoAvatar by +1.3 dB PSNR at up to 60x shorter per-subject schedule than common SOTA baselines. Website: https://spatialwalk.github.io/SpatialAvatar-0.

Problem

• 4D head avatar
• 3D Gaussian Splatting
• domain bias
• per-subject refinement
• representation sharing

Innovation

• 3D Gaussian Splatting
• FLAME-mesh-bound representation
• multi-stage reconstruction
• layout-preserving refinement
• zero-shot cross-domain avatar