Research map / Video / Video generation
FreeInit: Bridging Initialization Gap in Video Diffusion Models
Tianxing Wu, Chenyang Si, Yuming Jiang, Ziqi Huang, Ziwei 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 |
|---|---|---|---|
| apply FreeInit iterative low-frequency noise refinement at inference | AnimateDiff · UCF-101 prompts | DINO: 85.24 -> 92.01 | Improved |
| apply FreeInit | AnimateDiff · UCF-101 zero-shot | FVD: 1340.96 -> 1032.47 | Improved |
| apply FreeInit | ModelScope · UCF-101 prompts | Dynamic Degree |Delta UCF|: 3.71 -> 8.22 | Hurt |
| add noise reinitialization with Gaussian low-pass filter (ablation) | FreeInit on AnimateDiff · UCF-101 prompts | DINO: 86.77 -> 92.01 | Improved |
| use Gaussian instead of ideal low pass filter in reinitialization (ablation) | FreeInit on AnimateDiff · UCF-101 prompts | DINO: 87.53 -> 92.01 | Improved |
| increase FreeInit iterations (ablation) | FreeInit · UCF-101 and MSR-VTT | DINO | Improved |
| FreeInit with 2n steps versus vanilla DDIM with 5n steps (ablation) | FreeInit · UCF-101 | DINO | Improved |
| increase low-pass cut-off D0 from 0.25 to 0.5 (ablation) | FreeInit on AnimateDiff · UCF-101 prompts | Dynamic Degree |Delta UCF|: DINO 92.07 -> 93.91; DD diff 1.59 -> 37.78 | Mixed |
| apply FreeInit to videos with small fast-moving objects | FreeInit · failure case, waterdrops prompt | qualitative | Hurt |
Closest papers by meaning
- FreeNoise: Tuning-Free Longer Video Diffusion via Noise Rescheduling (ICLR 2024 · similarity 0.93)
- AR-Diffusion: Asynchronous Video Generation with Auto-Regressive Diffusion (CVPR 2025 · similarity 0.91)
- Ouroboros-Diffusion: Exploring Consistent Content Generation in Tuning-free Long Video Diffusion (AAAI 2025 · similarity 0.90)
- FreqPrior: Improving Video Diffusion Models with Frequency Filtering Gaussian Noise (ICLR 2025 · similarity 0.90)
- FreeLong: Training-Free Long Video Generation with SpectralBlend Temporal Attention (NeurIPS 2024 · similarity 0.90)
- FIFO-Diffusion: Generating Infinite Videos from Text without Training (NeurIPS 2024 · similarity 0.90)
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.
