Research map / Video / Video generation
GLOBER: Coherent Non-autoregressive Video Generation via GLOBal Guided Video DecodER
Mingzhen Sun, Weining Wang, Zihan Qin, Jiahui Sun, Sihan Chen, Jing Liu
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 CRA loss to video auto-encoder training (ablation) | GLOBER · UCF-101 | FVD: reduction of 37.0 | Improved |
| add CRA loss to video auto-encoder training (ablation) | GLOBER · Sky Time-lapse | FVD: reduction of 20.8 | Improved |
| add CRA loss to video auto-encoder training (ablation) | GLOBER · TaiChi-HD | FVD: reduction of 31.2 | Improved |
| global feature shape 8x8x64 -> 16x16x16 (ablation) | GLOBER · UCF-101 | FVD_gen: 725.1 -> 560.4 | Improved |
| global feature channels 16 -> 32 at 16x16 (ablation) | GLOBER · UCF-101 | FVD_gen: 560.4 -> 521.2 | Improved |
| unconditional guidance scale mu 0 -> 6 (reconstruction) (ablation) | GLOBER video decoder · UCF-101 256^2 | FVD: 211.4 -> 106.7 | Improved |
| unconditional guidance scale mu 6 -> 15 (reconstruction) (ablation) | GLOBER video decoder · UCF-101 256^2 | FVD: 106.7 -> 670.8 | Hurt |
| unconditional guidance scale mu 0 -> 12 (generation) (ablation) | GLOBER video generator · UCF-101 256^2 | FVD: 575.6 -> 168.9 | Improved |
| unconditional guidance scale mu 12 -> 15 (generation) (ablation) | GLOBER video generator · UCF-101 256^2 | FVD: 168.9 -> 224.3 | Hurt |
| non-autoregressive generation vs VIDM autoregressive, 128 frames | GLOBER · UCF-101 128-frame | FVD: 1531.9 -> 1177.4 | Improved |
| GLOBER vs VideoFusion sampling speed at F=64 | GLOBER · 256^2 sampling | time: 76s -> 21s | Improved |
| GLOBER vs StyleGAN-V on Sky Time-lapse | GLOBER · Sky Time-lapse | FVD: 79.5 -> 78.1 | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 6 of 6 on most shared results.
- VIDM — VIDM: Video Implicit Diffusion Models (AAAI 2023) · better on 5 of 5 shared results
- VideoFusion — VideoFusion: Decomposed Diffusion Models for High-Quality Video Generation (CVPR 2023) · better on 4 of 4 shared results
- StyleGAN-V — StyleGAN-V: A Continuous Video Generator With the Price, Image Quality and Perks of StyleGAN2 (CVPR 2022) · better on 4 of 4 shared results
- MOSO — MOSO: Decomposing MOtion, Scene and Object for Video Prediction (CVPR 2023) · better on 1 of 1 shared result
- Make-A-Video — Make-A-Video: Text-to-Video Generation without Text-Video Data (ICLR 2023) · better on 1 of 1 shared result
- CogVideo — CogVideo: Large-scale Pretraining for Text-to-Video Generation via Transformers (ICLR 2023) · better on 1 of 1 shared result
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
- Generative Pre-trained Autoregressive Diffusion Transformer (NeurIPS 2025 · similarity 0.88)
- Autoregressive Video Generation without Vector Quantization (ICLR 2025 · similarity 0.88)
- VideoMAR: Autoregressive Video Generation with Continuous Tokens (NeurIPS 2025 · similarity 0.87)
- InfinityStar: Unified Spacetime AutoRegressive Modeling for Visual Generation (NeurIPS 2025 · similarity 0.87)
- WorldWeaver: Generating Long-Horizon Video Worlds via Rich Perception (NeurIPS 2025 · similarity 0.87)
- From Slow Bidirectional to Fast Autoregressive Video Diffusion Models (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.
