Research map / Video / Video generation
Diffusion-based video editing
Edits video consistently with diffusion feature correspondence, reward tuning and frequency-aware factorization.
92 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2020–2026) · an approach in Video generation, keyword cluster “audio · editing · talking”
State History Tried Checklist Preprints Who Export
Where it stands
Not enough benchmark data Fewer than four papers report the same benchmark, so there is no record to follow.
32 papers in 2023–24 and 58 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic shrank, from 14.0% to 5.2%.
How this is decided
A benchmark is tracked when at least four papers of this approach report their own method on the same dataset and metric in at least two different years. Its record is the best number reported up to 2024 compared with the best reported in 2025–26: still improving means at least 1% better, flat means at most 0.2% better with at least two papers from 2025–26 trying. Numbers on a different scale from the rest (a fraction among percentages) and far outliers are left out. Results come from different papers with different setups, so they are signals, not head-to-head comparisons. The share of papers is counted at ICML, NeurIPS, ICLR, CVPR and AAAI.
How it got here
Most cited paper of each year
- 2020 · Consistent Video Style Transfer via Compound Regularization (AAAI 2020 · 45 citations)
- 2023 · Blind Video Deflickering by Neural Filtering With a Flawed Atlas (CVPR 2023 · 27 citations)
- 2024 · Video-P2P: Video Editing with Cross-attention Control (CVPR 2024 · 123 citations)
- 2025 · Go-with-the-Flow: Motion-Controllable Video Diffusion Models Using Real-Time Warped Noise (CVPR 2025 · 23 citations)
- 2026 · ContextFlow: Training-Free Video Object Editing via Adaptive Context Enrichment (AAAI 2026 · 4 citations)
Earliest papers here
- Consistent Video Style Transfer via Compound Regularization (AAAI 2020 · 45 citations)
- Non-Adversarial Video Synthesis With Learned Priors (CVPR 2020 · 16 citations)
- Blind Video Deflickering by Neural Filtering With a Flawed Atlas (CVPR 2023 · 27 citations)
What has been tried
57 papers here report 546 measured changes (a component added, removed or swapped, compared with a baseline): helped 75%, no clear effect 3%, hurt 20%, mixed 1%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| decoupled-guidance attention control (optimized embedding for source, initialized embedding for target) | DAVIS-based test set (45 samples) · Masked PSNR | 18.97 -> 20.65 | Helped | Video-P2P: Video Editing with Cross-attention Control (CVPR 2024 · 123 citations) |
| replace per-frame self-attention with VidToMe token merging | Video editing · Interpolation Error | 0.253 -> 0.105 | Helped | VidToMe: Video Token Merging for Zero-Shot Video Editing (CVPR 2024 · 45 citations) |
| propagate edited diffusion features using original-video correspondences instead of relying on extended attention alone | video editing · temporal consistency | Helped | TokenFlow: Consistent Diffusion Features for Consistent Video Editing (ICLR 2024 · 42 citations) | |
| Use pairwise SMM differences instead of full space-time feature reconstruction | Text-driven motion transfer · edit flexibility | Helped | Space-Time Diffusion Features for Zero-Shot Text-Driven Motion Transfer (CVPR 2024 · 36 citations) | |
| add flow-warped video to spatial controls | Condition combinations · winning rate against full FlowVid | 9% -> 38% | Helped | FlowVid: Taming Imperfect Optical Flows for Consistent Video-to-Video Synthesis (CVPR 2024 · 35 citations) |
| Add structure guidance for re-texturing | Re-texturing · structural fidelity | Helped | AVID: Any-Length Video Inpainting with Diffusion Model (CVPR 2024 · 34 citations) | |
| Add an edited key-frame appearance reference | BalanceCC user study · overall impression (MOS) | 2.88 -> 3.87 | Helped | CCEdit: Creative and Controllable Video Editing via Diffusion Models (CVPR 2024 · 31 citations) |
| Replace Consistent-Sparse Attention with Sparse Attention | motion editing videos · motion reliability | Hurt | MotionEditor: Editing Video Motion via Content-Aware Diffusion (CVPR 2024 · 28 citations) | |
| Add noise shuffling at each denoising step | 8-, 36-, and 90-frame videos · CLIP-F (×10−2) | 93.98 -> 95.95; 89.90 -> 93.18; 92.49 -> 95.99 | Helped | RAVE: Randomized Noise Shuffling for Fast and Consistent Video Editing with Diffusion Models (CVPR 2024 · 28 citations) |
| add warped-noise motion conditioning | DAVIS motion transfer I2V · FID | 76.62 -> 78.6 | Helped | Go-with-the-Flow: Motion-Controllable Video Diffusion Models Using Real-Time Warped Noise (CVPR 2025 · 23 citations) |
| Use BLIP instead of CLIP for video-text alignment | VE-Bench DB 10-fold · SROCC | 0.6379 -> 0.7171 | Helped | VE-Bench: Subjective-Aligned Benchmark Suite for Text-Driven Video Editing Quality Assessment (AAAI 2025 · 21 citations) |
| Add temporal-related adaptation | Video translation ablation · Pixel-MSE | 0.032 -> 0.015 | Helped | FRESCO: Spatial-Temporal Correspondence for Zero-Shot Video Translation (CVPR 2024 · 20 citations) |
What reviewers will expect
Datasets and baselines named in the 32 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- DAVIS · 12 of 32 papers
- WebVid-10M · 3 of 32 papers
- Pexels videos · 3 of 32 papers
- YouTube-VOS · 2 of 32 papers
- REDS · 2 of 32 papers
- Omni-Edit · 2 of 32 papers
Baselines
- TokenFlow · 7 of 32 papers
- AnyV2V · 6 of 32 papers
- FateZero · 5 of 32 papers
- Video-P2P · 5 of 32 papers
- ProPainter · 5 of 32 papers
- Flatten · 5 of 32 papers
- VACE · 4 of 32 papers
- STTN · 3 of 32 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 90 papers; from OpenAlex and the papers' affiliations).
- The Chinese University of Hong Kong · 9 papers
- The University of Hong Kong · 5 papers
- Adobe · 5 papers
- Meta · 5 papers
- Peking University · 4 papers
- Korea Advanced Institute of Science and Technology · 4 papers
- National University of Singapore · 4 papers
- ByteDance · 4 papers
- Nanyang Technological University · 3 papers
- NVIDIA · 3 papers
Use it
Download BibTeX of 25 key papersOpen the topic in the interactive map
Get alerts. Describe your idea in scime atlas and save it (free scime account): you get a weekly email when new preprints land close to it.
Related approaches
Same keyword cluster
- Diffusion-based human image animation (126 papers)
- Talking face generation (81 papers)
- Video-to-audio synthesis (76 papers)
- Text-to-motion diffusion (140 papers)
Elsewhere in the topic
- Efficient video diffusion models (283 papers)
- Trajectory-guided motion control (184 papers)
- Compositional text-to-video diffusion (159 papers)
- Autoregressive long video generation (156 papers)
- World-model video generation (110 papers)
Benchmark numbers, ablations and baseline numbers are extracted by AI from each paper's full text and can be wrong; the paper is the authority. Data: results index of 2026-11-02, citation counts from OpenAlex (2026-10-09).