Research map / Video / Video generation
TPDiff: Temporal Pyramid Video Diffusion Model
Lingmin Ran, Mike Zheng Shou
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 temporal pyramid (progressive frame rate) instead of full frame rate throughout | MiniFlux-vid · inference latency | latency (s): 20.79 -> 12.18 | Improved |
| Use temporal pyramid | AnimateDiff · inference latency | latency (s): 6.01 -> 4.04 | Improved |
| Use temporal pyramid with stage-wise diffusion | AnimateDiff and MiniFlux-vid · training convergence (FVD vs GPU hours) | training speed: 2x / 2.13x speedup | Improved |
| Use temporal pyramid | AnimateDiff · VBench | total score: 73.65% -> 74.96% | Improved |
| Use temporal pyramid | MiniFlux-vid · VBench | total score: 77.69% -> 78.52% | Improved |
| Use temporal pyramid | AnimateDiff · VBench | multiple objects: 39.42% -> 31.15% | Hurt |
| Use temporal pyramid | MiniFlux-vid · VBench | spatial relationship: 49.79% -> 42.23% | Hurt |
| Add data-noise alignment (ablation) | MiniFlux-vid · MSRVTT zero-shot | FVD: 602.1 -> 562.6 | Improved |
| Add data-noise alignment (ablation) | AnimateDiff · MSRVTT zero-shot | FVD: 834.5 -> 782.6 | Improved |
| Add renoising at stage transitions at inference (ablation) | TPDiff · qualitative | flicker | Improved |
| Model each stage as a complete DDIM process (ablation) | TPDiff with DDIM · training | convergence | Hurt |
Closest papers by meaning
- TRIP: Temporal Residual Learning with Image Noise Prior for Image-to-Video Diffusion Models (CVPR 2024 · similarity 0.90)
- AR-Diffusion: Asynchronous Video Generation with Auto-Regressive Diffusion (CVPR 2025 · similarity 0.90)
- Identifying and Solving Conditional Image Leakage in Image-to-Video Diffusion Model (NeurIPS 2024 · similarity 0.89)
- Fast and Memory-Efficient Video Diffusion Using Streamlined Inference (NeurIPS 2024 · similarity 0.89)
- Towards Precise Scaling Laws for Video Diffusion Transformers (CVPR 2025 · similarity 0.89)
- AsyncDiff: Parallelizing Diffusion Models by Asynchronous Denoising (NeurIPS 2024 · 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.
