Research map / Video / Video generation
LiON-LoRA: Rethinking LoRA Fusion to Unify Controllable Spatial and Temporal Generation for Video Diffusion
Yisu Zhang, Chenjie Cao, Chaohui Yu, Jianke Zhu
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 scaling token for camera/motion amplitude (ablation) | LiON-LoRA · Camera and motion strength control | linear controllability | Improved |
| add scaling token for motion strength instead of adapter scale (ablation) | LiON-LoRA · Motion strength (Pearson correlation with optical flow) | Pearson correlation | Improved |
| add layer-wise norm normalization of LoRA outputs (ablation) | LiON-LoRA fusion · Camera LoRA fusion (forward + orbit left) | fusion smoothness | Improved |
| increase training set from 100 to 7k samples at 4k iterations (ablation) | LiON-LoRA · Orbit left camera control | RotErr: 0.802 -> 1.227 | Hurt |
| increase training set from 100 to 7k samples at 4k iterations (ablation) | LiON-LoRA · Orbit left camera control | FVD: 175.3 -> 232.5 | Hurt |
| train 10k instead of 4k iterations with 100 samples (ablation) | LiON-LoRA · Orbit left camera control | RotErr: 0.802 -> 0.710 | Improved |
| train 10k instead of 4k iterations with 100 samples (ablation) | LiON-LoRA · Orbit left camera control | FVD: 175.3 -> 184.6 | Hurt |
| train 10k instead of 4k iterations with 7k samples (ablation) | LiON-LoRA · Orbit left camera control | RotErr: 1.227 -> 0.896 | Improved |
| LiON-LoRA vs DimensionX-S* on basic camera poses | CogVideoX · DL3DV basic camera poses | FVD: 193.3 -> 136.0 | Improved |
| LiON-LoRA vs CamI2V on complex fused poses | CogVideoX · DL3DV complex fused camera poses | RotErr: 0.924 -> 1.044 | Hurt |
Closest papers by meaning
- Video Diffusion Models are Training-free Motion Interpreter and Controller (NeurIPS 2024 · similarity 0.89)
- VD3D: Taming Large Video Diffusion Transformers for 3D Camera Control (ICLR 2025 · similarity 0.89)
- Wan-Move: Motion-controllable Video Generation via Latent Trajectory Guidance (NeurIPS 2025 · similarity 0.88)
- AC3D: Analyzing and Improving 3D Camera Control in Video Diffusion Transformers (CVPR 2025 · similarity 0.88)
- FloVD: Optical Flow Meets Video Diffusion Model for Enhanced Camera-Controlled 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)
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.
