Research map / Video / Video generation
SparseCtrl: Adding Sparse Controls to Text-to-Video Diffusion Models
Yuwei Guo, Ceyuan Yang, Anyi Rao, Maneesh Agrawala, Dahua Lin, Bo Dai
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 SparseCtrl sparse encoder instead of frame-wise ControlNet as control becomes sparser | AnimateDiff · Sparse depth-to-video, r_mask=7/8 | depth MAE err. (x100): 14.84 -> 7.40 | Improved |
| use SparseCtrl in dense setting (r_mask=0) compared with dense-control baselines | AnimateDiff · Sparse depth-to-video, r_mask=0 | depth MAE err. (x100): 8.24 -> 8.92 | Hurt |
| add temporal propagation layers to the frame-wise condition encoder (ablation) | SparseCtrl encoder · image animation (qualitative) | temporal consistency | Improved |
| remove noised sample input from the condition encoder (ablation) | SparseCtrl encoder with propagation layers · image animation (qualitative) | visual quality | Improved |
| use SparseCtrl instead of VideoComposer for image animation | image animation · Image animation (8 images) | LPIPS first-frame: 0.3346 -> 0.1467 | Improved |
| use SparseCtrl instead of DynamiCrafter for image animation | image animation · Image animation (8 images) | CLIP temporal consistency: 98.49 -> 95.25 | Hurt |
| use incomplete or uninformative text prompts with sketch condition (ablation) | SparseCtrl sketch-to-video · sketch-to-video (qualitative) | content generation | Hurt |
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.
- VideoComposer — VideoComposer: Compositional Video Synthesis with Motion Controllability (NeurIPS 2023) · better on 2 of 3 shared results
- Text2Video-Zero — Text2Video-Zero: Text-to-Image Diffusion Models are Zero-Shot Video Generators (ICCV 2023) · better on 0 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
- ControlVideo: Training-free Controllable Text-to-video Generation (ICLR 2024 · similarity 0.89)
- Encapsulated Composition of Text-to-Image and Text-to-Video Models for High-Quality Video Synthesis (CVPR 2025 · similarity 0.88)
- Ctrl-Adapter: An Efficient and Versatile Framework for Adapting Diverse Controls to Any Diffusion Model (ICLR 2025 · similarity 0.88)
- Enhancing Motion in Text-to-Video Generation with Decomposed Encoding and Conditioning (NeurIPS 2024 · similarity 0.88)
- Motion Prompting: Controlling Video Generation with Motion Trajectories (CVPR 2025 · similarity 0.87)
- Sparse Video-Gen: Accelerating Video Diffusion Transformers with Spatial-Temporal Sparsity (ICML 2025 · 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.
