Research map / Video / Video generation
What Are You Doing? A Closer Look at Controllable Human Video Generation
Emanuele Bugliarello, Anurag Arnab, Roni Paiss, Christy Koh, Pieter-Jan Kindermans, Cordelia Schmid
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 person-masked pICD instead of frame-wide ICD for person consistency (ablation) | human video evaluation · Side-by-side human study | agreement with human preference: 67.33 -> 72.67 | Improved |
| use FVD instead of FID for video quality (ablation) | metric choice · Side-by-side human study | Spearman correlation: 22.24 -> 96.36 | Improved |
| pAPE pose-adherence metric versus human judgment | pAPE · Side-by-side human study | pairwise accuracy: 71.95 | Improved |
| add both pose and text guidance to VACE (ablation) | VACE-Wan2.1 · WYD16 | video and human metrics | Improved |
| replace OpenPose with DWPose detector for MagicPose conditioning (ablation) | MagicPose · WYD16 | performance | Improved |
| more actors in the video (2 and 3+ versus 1) | MimicMotion, ControlNeXt, VACE · WYD16 | FVD and pAPE | Hurt |
| evaluate on WYD instead of TikTok/TED-Talks | pose-conditioned I2V models · human metrics | pICD and pAPE error: 1.8-4.6x higher pICD error; 1.8-12.3x higher pAPE error | Hurt |
| larger human actor size in frame | benchmarked models · WYD16 | pAPE vs FVD | Mixed |
Closest papers by meaning
- Video-Bench: Human-Aligned Video Generation Benchmark (CVPR 2025 · similarity 0.91)
- VBench: Comprehensive Benchmark Suite for Video Generative Models (CVPR 2024 · similarity 0.90)
- HumanVid: Demystifying Training Data for Camera-controllable Human Image Animation (NeurIPS 2024 · similarity 0.90)
- OpenHumanVid: A Large-Scale High-Quality Dataset for Enhancing Human-Centric Video Generation (CVPR 2025 · similarity 0.89)
- WorldModelBench: Judging Video Generation Models As World Models (NeurIPS 2025 · similarity 0.89)
- HumanDreamer: Generating Controllable Human-Motion Videos via Decoupled Generation (CVPR 2025 · 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.
