Research map / Video / Video segmentation and tracking
SimpleGVR: A Simple Baseline for Latent-Cascaded Generative Video Super-Resolution
Liangbin Xie, Yu Li, Shian Du, Menghan Xia, Xintao Wang, Fanghua Yu, Ziyan Chen, Pengfei Wan, Jiantao Zhou, Chao Dong
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 |
|---|---|---|---|
| replace end-to-end 1080p generation with cascaded 512p T2V plus SimpleGVR | large T2V model · AIGC100 | DOVER Overall: 62.32 -> 71.34 | Improved |
| replace end-to-end generation with cascaded pipeline | large T2V model · AIGC100 | VBench consistency and smoothness: Subject Consistency 95.16 -> 94.32 | No clear effect |
| replace swin attention with sparse local attention (ablation) | SimpleGVR · AIGC100, 17-frame | MUSIQ: 59.33 -> 61.64 | Improved |
| replace full attention with sparse local attention (ablation) | SimpleGVR · AIGC100, 17-frame | DOVER Overall: 69.64 -> 69.11 | No clear effect |
| replace full attention with sparse local attention (ablation) | SimpleGVR · AIGC100, 17-frame | FLOPS (G): 1612.5 -> 377.1 | Improved |
| add fine-tuning with flow-based and model-guided degradation (ablation) | SimpleGVR · AIGC100 | temporal consistency, color blending (qualitative) | Improved |
| replace uniform timestep sampler with detail-aware sampler (ablation) | SimpleGVR · AIGC100, 17-frame, 20K iterations | no-reference quality metrics | Improved |
| train with middle-range noise augmentation (0.3-0.6) instead of small or large (ablation) | SimpleGVR · AIGC100 (qualitative) | detail and structure correction | Improved |
| use small noise augmentation range 0.0-0.3 (ablation) | SimpleGVR · AIGC100 (qualitative) | artifact removal | Hurt |
| use large noise augmentation range 0.6-0.9 (ablation) | SimpleGVR · AIGC100 (qualitative) | structure preservation | Hurt |
| replace full attention with swin attention (ablation) | SimpleGVR · AIGC100, 17-frame | MUSIQ: 62.19 -> 59.33 | Hurt |
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.
- FlashVideo — FlashVideo: Flowing Fidelity to Detail for Efficient High-Resolution Video Generation (AAAI 2026) · better on 4 of 4 shared results
- Upscale-A-Video — Upscale-A-Video: Temporal-Consistent Diffusion Model for Real-World Video Super-Resolution (CVPR 2024) · better on 2 of 2 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
- TurboVSR: Fantastic Video Upscalers and Where to Find Them (ICCV 2025 · similarity 0.90)
- VideoGigaGAN: Towards Detail-rich Video Super-Resolution (CVPR 2025 · similarity 0.90)
- PatchVSR: Breaking Video Diffusion Resolution Limits with Patch-wise Video Super-Resolution (CVPR 2025 · similarity 0.88)
- Upscale-A-Video: Temporal-Consistent Diffusion Model for Real-World Video Super-Resolution (CVPR 2024 · similarity 0.88)
- Arbitrary-Scale Video Super-resolution Guided by Dynamic Context (AAAI 2024 · similarity 0.88)
- NegVSR: Augmenting Negatives for Generalized Noise Modeling in Real-world Video Super-Resolution (AAAI 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.
