WildActor: Unconstrained Identity-Preserving Video Generation

2026-02-28

AI Summary

Existing human video generation methods struggle to maintain full-body identity consistency under multi-view settings, dynamic camera motions, and large-scale body movements, often suffering from facial centralization or rigid poses. To address this, this work proposes WildActor, a novel framework that integrates an asymmetric identity-preserving attention mechanism and a viewpoint-adaptive Monte Carlo sampling strategy. By leveraging tri-view representation learning and margin-based utility-aware reweighting of reference conditions, WildActor enables high-fidelity, identity-consistent synthesis from arbitrary viewpoints. Furthermore, we introduce Actor-18M, the first large-scale human video dataset supporting unconstrained camera angles and environments, and establish Actor-Bench, a new benchmark on which WildActor demonstrates significant superiority over existing approaches in diverse shot compositions, extreme viewpoint transitions, and complex motion scenarios.

Technology Category

• Computer Vision: Video Understanding & Activity Analysis
• Humans and AI: Human-Aware Planning and Behavior Prediction
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Economics, Online Markets and Human Computation: Economic ramifications for generative AI infrastructure and applications

Abstract

Production-ready human video generation requires digital actors to maintain strictly consistent full-body identities across dynamic shots, viewpoints and motions, a setting that remains challenging for existing methods. Prior methods often suffer from face-centric behavior that neglects body-level consistency, or produce copy-paste artifacts where subjects appear rigid due to pose locking. We present Actor-18M, a large-scale human video dataset designed to capture identity consistency under unconstrained viewpoints and environments. Actor-18M comprises 1.6M videos with 18M corresponding human images, covering both arbitrary views and canonical three-view representations. Leveraging Actor-18M, we propose WildActor, a framework for any-view conditioned human video generation. We introduce an Asymmetric Identity-Preserving Attention mechanism coupled with a Viewpoint-Adaptive Monte Carlo Sampling strategy that iteratively re-weights reference conditions by marginal utility for balanced manifold coverage. Evaluated on the proposed Actor-Bench, WildActor consistently preserves body identity under diverse shot compositions, large viewpoint transitions, and substantial motions, surpassing existing methods in these challenging settings.

Problem

• identity-preserving
• video generation
• full-body consistency
• unconstrained viewpoints
• human video synthesis

Innovation

• Identity-Preserving Video Generation
• Asymmetric Identity-Preserving Attention
• Viewpoint-Adaptive Monte Carlo Sampling
• Actor-18M Dataset
• Any-View Human Synthesis