Research map / Video / Video generation
FreeViS: Training-free Video Stylization with Inconsistent References
Jiacong Xu, Yiqun Mei, Ke Zhang, Vishal M. Patel
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use multiple stylized references instead of a single first-frame reference (vs AnyV2V* re-implementation) | FreeViS · Video style transfer (200 videos) | Style Consistency: 0.675 -> 0.898 | Improved |
| use FreeViS instead of AnyV2V* on the same base model | I2V stylization · Video style transfer (200 videos) | Human preference: 3.372 -> 4.113 | Improved |
| use FreeViS instead of AnyV2V* | I2V stylization · Video style transfer (200 videos) | Flow Consistency (EPE): 1.103 -> 0.641 | Improved |
| use FreeViS instead of AnyV2V* | I2V stylization · Video style transfer (200 videos) | LPIPS: 0.218 -> 0.479 | No clear effect |
| use FreeViS instead of AnyV2V* | I2V stylization · Video style transfer (200 videos) | FID: 27.59 -> 21.62 | Improved |
| add IHC (high-frequency compensation) for layout reconstruction (ablation) | FreeViS · Video style transfer (200 videos) | qualitative layout fidelity | Improved |
| add extra reference frames (ablation) | FreeViS · Video style transfer (200 videos) | qualitative style consistency | Improved |
| add EOG (flow-based motion cues) for plain regions (ablation) | FreeViS · Video style transfer (200 videos) | flow consistency and stylization quality | Improved |
| apply FreeViS on Wan2.1 videos instead of StyleCrafter | stylized T2V · Stylized T2V generation (200 prompts) | Dynamic Quality: 0.368 -> 0.509 | Improved |
| apply FreeViS on Wan2.1 videos instead of StyleCrafter | stylized T2V · Stylized T2V generation (200 prompts) | CSD Score: 0.515 -> 0.437 | Hurt |
| apply FreeViS on top of AnyV2V* implementation | FreeViS · Video style transfer (200 videos) | inference time: approximately 30% overhead | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- StyleMaster — StyleMaster: Stylize Your Video with Artistic Generation and Translation (CVPR 2025) · better on 3 of 3 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- DreamStyle: A Unified Framework for Video Stylization (CVPR 2026 · similarity 0.91)
- StyleMaster: Stylize Your Video with Artistic Generation and Translation (CVPR 2025 · similarity 0.90)
- Local Consistency Guidance: Personalized Stylization Method of Face Video (Student Abstract) (AAAI 2024 · similarity 0.87)
- Generative Human Motion Stylization in Latent Space (ICLR 2024 · similarity 0.87)
- Styl3R: Instant 3D Stylized Reconstruction for Arbitrary Scenes and Styles (NeurIPS 2025 · similarity 0.87)
- FreePCA: Integrating Consistency Information across Long-short Frames in Training-free Long Video Generation via Principal Component Analysis (CVPR 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
