Synthetic Video Enhances Physical Fidelity in Video Synthesis
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866913760774455296 |
|---|---|
| author | Zhao, Qi Ni, Xingyu Wang, Ziyu Cheng, Feng Yang, Ziyan Jiang, Lu Wang, Bohan |
| author_facet | Zhao, Qi Ni, Xingyu Wang, Ziyu Cheng, Feng Yang, Ziyan Jiang, Lu Wang, Bohan |
| contents | We investigate how to enhance the physical fidelity of video generation models by leveraging synthetic videos derived from computer graphics pipelines. These rendered videos respect real-world physics, such as maintaining 3D consistency, and serve as a valuable resource that can potentially improve video generation models. To harness this potential, we propose a solution that curates and integrates synthetic data while introducing a method to transfer its physical realism to the model, significantly reducing unwanted artifacts. Through experiments on three representative tasks emphasizing physical consistency, we demonstrate its efficacy in enhancing physical fidelity. While our model still lacks a deep understanding of physics, our work offers one of the first empirical demonstrations that synthetic video enhances physical fidelity in video synthesis. Website: https://kevinz8866.github.io/simulation/ |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2503_20822 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Synthetic Video Enhances Physical Fidelity in Video Synthesis Zhao, Qi Ni, Xingyu Wang, Ziyu Cheng, Feng Yang, Ziyan Jiang, Lu Wang, Bohan Image and Video Processing Artificial Intelligence Graphics We investigate how to enhance the physical fidelity of video generation models by leveraging synthetic videos derived from computer graphics pipelines. These rendered videos respect real-world physics, such as maintaining 3D consistency, and serve as a valuable resource that can potentially improve video generation models. To harness this potential, we propose a solution that curates and integrates synthetic data while introducing a method to transfer its physical realism to the model, significantly reducing unwanted artifacts. Through experiments on three representative tasks emphasizing physical consistency, we demonstrate its efficacy in enhancing physical fidelity. While our model still lacks a deep understanding of physics, our work offers one of the first empirical demonstrations that synthetic video enhances physical fidelity in video synthesis. Website: https://kevinz8866.github.io/simulation/ |
| title | Synthetic Video Enhances Physical Fidelity in Video Synthesis |
| topic | Image and Video Processing Artificial Intelligence Graphics |
| url | https://arxiv.org/abs/2503.20822 |