Research map / Video / Video generation
LaVie: High-Quality Video Generation with Cascaded Latent Diffusion Models
Yaohui Wang, Xinyuan Chen, Xin Ma, Shangchen Zhou, Ziqi Huang, Ceyuan Yang, Ceyuan Yang, Yinan He, Jiashuo Yu, Peiqing Yang, Yuwei Guo, Tianxing Wu and 8 more
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 joint image-video fine-tuning instead of fine-tuning the entire UNet on video only (ablation) | LaVie base T2V · qualitative training scheme comparison | visual quality / concept retention | Improved |
| use joint image-video fine-tuning instead of training only temporal modules (ablation) | LaVie base T2V · qualitative training scheme comparison | visual quality | Improved |
| train on WebVid10M+Vimeo25M and use class names as prompts, compared with Video LDM | LaVie · UCF101 zero-shot | FVD: 550.61 -> 526.30 | Improved |
| compare to Make-A-Video trained on larger HD-VILA-100M data with template prompts | LaVie · UCF101 zero-shot | FVD: 367.23 -> 526.30 | Hurt |
| compare to Make-A-Video on MSR-VTT | LaVie · MSR-VTT zero-shot | CLIPSIM: 0.3049 -> 0.2949 | Hurt |
| compare to Video LDM and ModelScope on MSR-VTT | LaVie · MSR-VTT zero-shot | CLIPSIM: 0.2929 -> 0.2949 | No clear effect |
| compare LaVie to ModelScope in human pairwise preference | LaVie · human preference overall quality | preference rate: 75.00% | Improved |
| compare LaVie to ModelScope on motion smoothness | LaVie · human evaluation | proportion good: 0.31 -> 0.35 | Improved |
| compare LaVie to ModelScope on background consistency | LaVie · human evaluation | proportion good: 0.63 -> 0.72 | Improved |
| compare LaVie to ModelScope on face/body/hand quality | LaVie · human evaluation | proportion good: 0.18 -> 0.24 | Improved |
Later papers that compared with this method
14 later accepted papers report LaVie as a baseline in their results tables; 12 of them report a better number on most of the results they share with it.
- MoSA: Motion-Coherent Human Video Generation via Structure-Appearance Decoupling (ICLR 2026 · better on 2 of 2 shared results)
- PhysRAG: Enhancing Physics-Awareness in Video Generation via Retrieval-Augmented Generation (ECCV 2026 · better on 1 of 1 shared result)
- Consistent and Controllable Image Animation With Linear Motion Diffusion Transformers (TPAMI 2026 · better on 1 of 1 shared result)
- VideoElevator: Elevating Video Generation Quality with Versatile Text-to-Image Diffusion Models (AAAI 2025 · better on 8 of 9 shared results)
- Optical-Flow Guided Prompt Optimization for Coherent Video Generation (CVPR 2025 · better on 5 of 6 shared results)
- VideoGuide: Improving Video Diffusion Models without Training Through a Teacher's Guide (CVPR 2025 · better on 3 of 3 shared results)
- REDUCIO! Generating 1K Video within 16 Seconds using Extremely Compressed Motion Latents (ICCV 2025 · better on 3 of 3 shared results)
- MagicTime: Time-Lapse Video Generation Models as Metamorphic Simulators (TPAMI 2025 · better on 3 of 3 shared results)
6 more not listed.
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 2 on most shared results.
- Make-A-Video — Make-A-Video: Text-to-Video Generation without Text-Video Data (ICLR 2023) · better on 0 of 2 shared results
- VideoFusion — VideoFusion: Decomposed Diffusion Models for High-Quality Video Generation (CVPR 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
- Encapsulated Composition of Text-to-Image and Text-to-Video Models for High-Quality Video Synthesis (CVPR 2025 · similarity 0.90)
- VideoElevator: Elevating Video Generation Quality with Versatile Text-to-Image Diffusion Models (AAAI 2025 · similarity 0.90)
- VideoCrafter2: Overcoming Data Limitations for High-Quality Video Diffusion Models (CVPR 2024 · similarity 0.90)
- Efficient Video Diffusion Models via Content-Frame Motion-Latent Decomposition (ICLR 2024 · similarity 0.90)
- Upscale-A-Video: Temporal-Consistent Diffusion Model for Real-World Video Super-Resolution (CVPR 2024 · similarity 0.90)
- VSTAR: Generative Temporal Nursing for Longer Dynamic Video Synthesis (ICLR 2025 · similarity 0.89)
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.
