FFAvatar: Few-Shot, Feed-Forward, and Generalizable Avatar Reconstruction

2026-05-14

AI Summary

This work addresses the limitations of traditional 3D head avatar reconstruction, which often relies on time-consuming optimization or costly preprocessing and thus lacks scalability. The authors propose a unified feed-forward framework capable of rapidly generating high-quality, animatable 3D Gaussian head avatars from only a few pose-free portrait images. Key innovations include the first Multi-View Query Transformer for fusing multi-image information, end-to-end regression of FLAME animation parameters, and a three-stage curriculum training strategy—comprising large-scale monocular pretraining, multi-view fine-tuning, and optional personalization—to balance generalization and fidelity. On the NeRSemble benchmark, the method surpasses the state-of-the-art LAM by 5.5 dB in PSNR, achieves reconstruction in just 2 seconds (10 seconds with personalization), and enables real-time animation at 49 FPS on a single A100 GPU.

Abstract

Avatar reconstruction has traditionally relied on per-subject optimization that requires hours of computation or on expensive preprocessing that limits scalability. We introduce FFAvatar, a generalizable feed-forward framework that reconstructs high-quality, animatable 3D Gaussian head avatars from few-shot unposed portrait images in seconds. FFAvatar fuses information from multiple source images into a unified canonical Gaussian representation through Multi-View Query-Former, which is animated via FLAME parameters predicted end-to-end directly from pixels, eliminating the overhead of offline FLAME extraction. We further propose a three-stage training curriculum that achieves both broad generalization and high-fidelity reconstruction: (i) scalable pretraining on extensive monocular video data with over 1M identities to learn strong generalizable priors; (ii) multi-view fine-tuning on a small but high-quality dataset of 360-degree captures to enhance geometric fidelity and extreme-view awareness; and (iii) optional personalization that adapts to specific identities for maximum fidelity within 500 optimization steps. Extensive experiments demonstrate that FFAvatar sets a new standard for identity preservation, geometric consistency, and animation fidelity. On the NeRSemble benchmark, it outperforms the state-of-the-art LAM by a substantial 5.5 PSNR gain. Furthermore, FFAvatar enables real-time deployment, reconstructing avatars in 2 seconds without personalization and 10 seconds with personalization, while supporting 49 FPS animation on a single NVIDIA A100 GPU.

Problem

• avatar reconstruction
• few-shot
• generalizable
• feed-forward
• 3D Gaussian

Innovation

• Few-Shot Avatar Reconstruction
• Feed-Forward Framework
• 3D Gaussian Avatars
• Multi-View Query-Former
• End-to-End FLAME Prediction