Research map / Video / Video generation
Bitrate-Controlled Diffusion for Disentangling Motion and Content in Video
Xiao Li, Qi Chen, Xiulian Peng, Kai Yu, Xie Chen, Yan Lu
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 |
|---|---|---|---|
| add cross-driven training strategy (ablation) | BCD · LRS3 motion transfer | cross error (x10^-2): 40.4 -> 3.67 | Improved |
| add cross-driven training strategy (ablation) | BCD · LRS3 motion transfer | FID: 120.1 -> 86.0 | Improved |
| add bitrate loss (ablation) | BCD · LRS3 motion transfer | cross error (x10^-2): 6.21 -> 3.67 | Improved |
| add bitrate loss (ablation) | BCD · LRS3 motion transfer | FID: 92.7 -> 86.0 | Improved |
| lower target bitrate from 4.0 to 2.0 kbps (ablation) | BCD · LRS3 motion transfer | cross error (x10^-2): 3.67 -> 5.68 | Hurt |
| raise target bitrate from 4.0 to 6.0 kbps (ablation) | BCD · LRS3 motion transfer | cross error (x10^-2): 3.67 -> 4.13 | Hurt |
| raise target bitrate from 4.0 to 8.0 kbps (ablation) | BCD · LRS3 motion transfer | cross error (x10^-2): 3.67 -> 3.74 | Mixed |
| use a single content reference frame instead of all frames (ablation) | BCD · LRS3 motion transfer | cross error (x10^-2): 3.67 -> 3.81 | Hurt |
| use 5 content frames instead of all (ablation) | BCD · LRS3 motion transfer | cross error (x10^-2): 3.67 -> 4.21 | Hurt |
| compare BCD with FOMM on motion error | BCD vs FOMM · LRS3 motion transfer | motion error (x10^-2): 24.3 -> 3.13 | Improved |
| compare BCD with LivePortrait on FID | BCD vs LivePortrait · LRS3 motion transfer | FID: 100.3 -> 86.0 | Improved |
| compare BCD with C-DSVAE on attribute accuracy | BCD vs C-DSVAE · Sprites | attribute accuracy: 100.00 -> 100.00 | No clear effect |
Closest papers by meaning
- Motion Consistency Model: Accelerating Video Diffusion with Disentangled Motion-Appearance Distillation (NeurIPS 2024 · similarity 0.88)
- Video Motion Transfer with Diffusion Transformers (CVPR 2025 · similarity 0.88)
- Efficient Video Diffusion Models via Content-Frame Motion-Latent Decomposition (ICLR 2024 · similarity 0.87)
- Emergent Temporal Correspondences from Video Diffusion Transformers (NeurIPS 2025 · similarity 0.87)
- MotionAura: Generating High-Quality and Motion Consistent Videos using Discrete Diffusion (ICLR 2025 · similarity 0.87)
- TTVidT: Decoupling the Temporal Axis for Efficient Motion-Centric Video Pretraining (NeurIPS 2026 · 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.
