Icon

Omni-Effects

Unified and Spatially-Controllable Visual Effects Generation

Fangyuan Mao* † Aiming Hao† Jintao Chen Dongxia Liu Xiaokun Feng Jiashu Zhu Meiqi Wu Chubin Chen Jiahong Wu‡ Xiangxiang Chu

*Work done during the internship at AMAP, Alibaba Group,†Equal contribution, ‡Project lead

AMAP, Alibaba Group

We're hiring video generation full-time employees and interns, contact us: hongxi.wjh@alibaba-inc.com

Research Paper GitHub Omni-Effects Omni-VFX

VFX Created from Omni-Effects

Prompt-guided VFX
Omni-Effects generates single-VFX results with fidelity and reliability based on the prompt-guided.
"Change the setting to spring with blooming flowers."
"Add fireworks."
"Change the setting to autumn with falling leaves."
"Turn characters into hulk."
"Super saiyan transformation, real life style."
"Character's face transforms into a metallic, robotic visage."
"Turn it into anime style."
"Turn it into impasto style."
"Turn it into mosaic style."
Prompt-guided Multi-VFX
Omni-Effects generates multi-VFX results with fidelity and reliability based on the prompt-guided.
"Transform into transformers, Change the setting to a winter scene."
"The character's eyes begin to glow blue, and a bright blue electric effect emanates from his body, Add fireworks."
"Character is surrounded by puppies, Turn it into 3D cartoon style."
SPA-guided VFX
Omni-Effects generates single-VFX results with fidelity and reliability based on the SPA-guided.
"Levitate left dog."
"Levitate right woman."
"Levitate right apple."
"Melt left cup."
"Melt right bottle."
"Melt right car."
"Explode right cup."
"Explode left coconut."
"Explode left lens."
SPA-guided Multi-VFX
Omni-Effects generates multi-VFX results with fidelity and reliability based on the SPA-guided.
"Melt left car; Levitate right car."
"Levitate left woman; Melt right dog."
"Melt left chair; Levitate right chair."
"Explode left ring; Melt right ring."
"Explode left car; Melt right car."
"Melt left bird; Explode right bird."

Ethics Concerns

The images used in these demos are sourced from public domains or generated by models, and are intended solely to showcase the capabilities of this research. If you have any concerns, please contact us at haoaiming.ham@alibaba-inc.com, and we will promptly remove them.

Acknowledgements

Our work is primarily based on CogVideoX, VFXCreator, Switch Transformers, EasyControl. We are sincerely grateful for their excellent works.

BibTeX

@misc{mao2025omnieffects,
          title={Omni-Effects: Unified and Spatially-Controllable Visual Effects Generation}, 
          author={Fangyuan Mao and Aiming Hao and Jintao Chen and Dongxia Liu and Xiaokun Feng and Jiashu Zhu and Meiqi Wu and Chubin Chen and Jiahong Wu and Xiangxiang Chu},
          year={2025},
          eprint={2508.07981},
          archivePrefix={arXiv},
          primaryClass={cs.CV}
      }