Research map / Video / Video generation
DiT as Real-Time Rerenderer: Streaming Video Stylization with Autoregressive Diffusion Transformer
Hengye Lyu, Zisu Li, Yue Hong, Yueting Weng, Jiaxin Shi, Hanwang Zhang, Chen Liang
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 |
|---|---|---|---|
| add Self Forcing + DMD post-training to the causal student (ablation) | RTR-DiT causal model · TV2V (50 Pexels videos) | CLIP-T: 0.2284 -> 0.2585 | Improved |
| add Self Forcing + DMD post-training to the causal student (ablation) | RTR-DiT causal model · RV2V (50 Pexels videos) | CSD-Score: 0.6002 -> 0.7958 | Improved |
| add post-training to the causal student (ablation) | RTR-DiT causal model · TV2V (50 Pexels videos) | Aesthetic Quality: 0.5446 -> 0.6302 | Improved |
| add post-training to the causal student (ablation) | RTR-DiT causal model · RV2V (50 Pexels videos) | Imaging Quality: 0.6021 -> 0.7321 | Improved |
| add post-training to the causal student (ablation) | RTR-DiT causal model · TV2V and RV2V | CLIP-F: 0.9948 -> 0.9914 (TV2V); 0.9956 -> 0.9909 (RV2V) | Hurt |
| distill bidirectional teacher into 2-step causal student (ablation) | RTR-DiT · TV2V (50 Pexels videos) | generation time (min): 1.43 -> 0.12 | Improved |
| distill bidirectional teacher into causal student (ablation) | RTR-DiT · RV2V (50 Pexels videos) | CSD-Score: 0.8050 -> 0.7958 | No clear effect |
| reference-preserving KV cache update (anchor reference at cache front) (ablation) | RTR-DiT · long RV2V videos | qualitative style consistency | Improved |
| use RTR-DiT instead of open-source TV2V baselines | RTR-DiT · TV2V (50 Pexels videos) | CLIP-T: 0.2122 -> 0.2585 (vs TokenWarping) | Improved |
| use RTR-DiT instead of commercial Gen-4 Aleph | RTR-DiT · RV2V (50 Pexels videos) | CSD-Score: 0.8312 -> 0.7958 | Hurt |
| use RTR-DiT instead of Gen-4 Aleph | RTR-DiT · RV2V (50 Pexels videos) | Aesthetic Quality: 0.6336 -> 0.6714 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- StyleMaster — StyleMaster: Stylize Your Video with Artistic Generation and Translation (CVPR 2025) · better on 3 of 3 shared results
- FRESCO — FRESCO: Spatial-Temporal Correspondence for Zero-Shot Video Translation (CVPR 2024) · 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
- From Slow Bidirectional to Fast Autoregressive Video Diffusion Models (CVPR 2025 · similarity 0.89)
- StreamingT2V: Consistent, Dynamic, and Extendable Long Video Generation from Text (CVPR 2025 · similarity 0.87)
- VideoElevator: Elevating Video Generation Quality with Versatile Text-to-Image Diffusion Models (AAAI 2025 · similarity 0.87)
- VMC: Video Motion Customization using Temporal Attention Adaption for Text-to-Video Diffusion Models (CVPR 2024 · similarity 0.87)
- Motion Consistency Model: Accelerating Video Diffusion with Disentangled Motion-Appearance Distillation (NeurIPS 2024 · similarity 0.87)
- Upscale-A-Video: Temporal-Consistent Diffusion Model for Real-World Video Super-Resolution (CVPR 2024 · similarity 0.87)
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.
