Abstract
Recent advances in 3D generation have enabled impressive visual synthesis, yet existing methods often rely on 2D diffusion supervision without explicit mechanisms for geometric consistency, leading to spatial hallucinations like duplicated structures and misaligned geometry. These issues are exacerbated in 4D generation, where maintaining consistency across viewpoints and temporal evolution introduces additional challenges, including jitter, identity flicker, and structural drift.
We present Hallo4D, a unified and model-agnostic framework for mitigating spatiotemporal hallucinations in 3D and 4D content generation. Hallo4D introduces a generation-detection-correction paradigm that leverages large multimodal language models (LMMs) to identify and summarize spatial and temporal inconsistencies from multi-view and multi-frame renderings. These insights guide a consensus-driven image-space consistency optimization, where an LMM-based selector evaluates candidate corrections through multi-model voting, without requiring retraining or architectural modifications.
To further enhance temporal consistency and optimization efficiency, Hallo4D incorporates motion-aware keyframe sampling, LMM-guided initialization, and appearance alignment. Additionally, we introduce exposure-aware optimization and visibility pruning to improve robustness under challenging viewpoints. Extensive experiments demonstrate that Hallo4D consistently outperforms strong baselines across diverse 3D and 4D generation settings, providing a scalable and generalizable solution for consistency-aware content generation.
Blogger's Review: Hallo4D effectively tackles consistency issues in 3D and 4D generation through the integration of multimodal language models, showcasing its strong technical potential and application prospects. This framework not only enhances generation quality but also offers new insights for future research. The consensus-driven approach demonstrates superiority in complex generation tasks, warranting further exploration and application.