EasyVFX: Frequency-Driven Decoupling for Resource-Efficient VFX Generation

Generating high-fidelity visual effects (VFX) typically demands massive datasets and prohibitive computational power due to the intricate coupling of spatial textures and temporal dynamics. In this paper, we introduce EasyVFX, a resource-efficient framework that achieves realistic VFX synthesis under stringent constraints. Our core philosophy lies in frequency-domain decomposition: we observe that the complexity of VFX can be significantly mitigated by decoupling high-frequency components, which represent intricate spatial appearances, from low-frequency components that encapsulate global motion dynamics. This spectral disentanglement transforms a high-dimensional learning problem into manageable sub-tasks, thereby lowering the optimization barrier and reducing data dependency. Building upon this insight, we propose a two-stage training paradigm. First, we design a Frequency-aware Mixture-of-Experts (MoE) architecture. By utilizing a Soft-MoE routing mechanism, our model assigns specialized experts to distinct spectral bands, enabling them to independently cultivate robust priors for appearance and motion. This specialization allows the model to acquire foundational VFX knowledge with only 8 GPUs and 5,000 training steps. Second, we introduce a Test-Time Training (TTT) strategy powered by a novel Frequency-Perturbation Loss. This allows the pre-trained model to swiftly adapt to specific, unseen effects through localized optimizations, requiring only 100 steps on a single GPU. Experimental results demonstrate that EasyVFX produces structurally consistent and visually stunning effects, proving that frequency-aware learning is a key catalyst for democratizing professional-grade VFX generation.

Comparison with state-of-the-art visual effect generation methods

Reference

Ours

CogVideoX

Omini-Effect

VFXCreator

Video as prompt

More Results

@inproceedings{ma2026easyvfx,
  title     = {EasyVFX: Frequency-Driven Decoupling for Resource-Efficient VFX Generation},
  author    = {Ma, Yue and Ye, Xu and Wang, Qinghe and Wang, Yucheng and Liu, Hongyu and Zhang, Yinhan and Wang, Xinyu and Che, Yuanpeng and Mo, Shanhui and Liang, Paul and Zhan, Fangneng and Chen, Qifeng},
  booktitle = {ACM SIGGRAPH 2026 Conference Proceedings},
  year      = {2026}
}