Research map / Video / Video generation
VideoMaMa++: Temporally Consistent Video Matting via Preserve-and-Refine Tokens
Sangbeom Lim, Seoung Wug Oh, Heeji Yoon, Seungryong Kim, Joon-Young Lee
Read the paperCheck your idea against 137,217 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Proposes VideoMaMa++, a diffusion-based video matting method that uses preserve-and-refine tokens to keep temporal consistency across independently processed chunks of frames.
Problem
Diffusion-based video matting is limited to short chunks of 6-24 frames due to heavy compute, and chunks are matted independently, causing temporal inconsistency across chunk boundaries for ordinary-length videos.
Approach
Introduces preserve-and-refine tokens to carry information across independently matted video chunks so that the matte stays temporally consistent; the abstract only partially describes this mechanism.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add reference anchor to sliding-window inference (ablation) | VideoMaMa++ · VideoMatte (1920x1080) | Grad: 2.2398 -> 1.7887 | Improved |
| add preserve token on overlap frames (ablation) | VideoMaMa++ · Chunk boundary overlap discrepancy | MAD: 0.7128 -> 0.2583 | Improved |
| use token concatenation instead of channel concatenation of RGB, mask and noise (ablation) | VideoMaMa++ · V-HIM60 (Hard) | MAD: 2.2253 -> 1.6924 | Improved |
| condition on SAM logits instead of binary masks (ablation) | VideoMaMa++ · V-HIM60 (Hard) | MAD: 1.6942 -> 1.6272 | Mixed |
| increase denoising steps from 1 to 5 at inference (ablation) | VideoMaMa++ · V-HIM60 (Hard) | MAD: 1.6680 -> 10.7571 | Hurt |
| use chunk size 6 instead of 3 (ablation) | VideoMaMa++ · V-HIM60 (Hard) | MAD: 2.1206 -> 1.6942 | Improved |
| use chunk size 12 instead of 6 (ablation) | VideoMaMa++ · V-HIM60 (Hard) | MAD: 1.6942 -> 1.7906 | Hurt |
| increase LoRA rank from 64 to 512 (ablation) | VideoMaMa++ · V-HIM60 (Hard) | MAD: 2.4294 -> 1.6942 | Improved |
| use random or last frame of first chunk as anchor instead of first frame (ablation) | VideoMaMa++ · V-HIM60 (Hard) | MAD: 1.69 -> 1.70 | No clear effect |
| refresh anchor every 3 chunks (ablation) | VideoMaMa++ · V-HIM60 (Hard) | MAD: 1.69 -> 1.70 | No clear effect |
| corrupt the anchor frame (full occlusion) (ablation) | VideoMaMa++ · V-HIM60 (Hard) | MAD: 1.71 -> 1.76 | Hurt |
| repeat interactive refinement iterations with preserve and refine tokens | VideoMaMa++ · V-HIM60 (Hard) | MAD: 1.6924 -> 1.6634 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- MatAnyone 2 — MatAnyone 2: Scaling Video Matting via a Learned Quality Evaluator (CVPR 2026) · better on 3 of 3 shared results
- MaGGIe — MaGGIe: Masked Guided Gradual Human Instance Matting (CVPR 2024) · better on 2 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.
Experimental setup
Builds on: VideoMaMa
Closest papers by meaning
- VideoMaMa: Mask-Guided Video Matting via Generative Prior (CVPR 2026 · similarity 0.94)
- Static-to-Dynamic: Animating Still Mattes via Generative Motion for Video Matting (NeurIPS 2026 · similarity 0.91)
- $\alpha$Matte4K & $\mu$Matting: Dataset and Model for Ultra-Micro Precision Alpha Video Matting (CVPR 2026 · similarity 0.91)
- MatAnyone: Stable Video Matting with Consistent Memory Propagation (CVPR 2025 · similarity 0.90)
- MaGGIe: Masked Guided Gradual Human Instance Matting (CVPR 2024 · similarity 0.89)
- DRIP: Unleashing Diffusion Priors for Joint Foreground and Alpha Prediction in Image Matting (NeurIPS 2024 · similarity 0.88)
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.
