Research map / Video / Video generation
A Good Image Generator Is What You Need for High-Resolution Video Synthesis
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 latent residual trajectories instead of hidden states directly as latent codes (ablation) | MoCoGAN-HD · UCF-101 | IS: 28.20 -> 33.95 | Helped |
| Use latent residual trajectories instead of hidden states directly as latent codes (ablation) | MoCoGAN-HD · UCF-101 | FVD: 790.87 -> 700.00 | Helped |
| Add the contrastive image discriminator (ablation) | MoCoGAN-HD · UCF-101 | FVD: 796.67 -> 700.00 | Helped |
| Add the video discriminator (ablation) | MoCoGAN-HD · UCF-101 | FVD: 867.43 -> 700.00 | Helped |
| Generate at 256×256 rather than 128×128 resolution (ablation) | MoCoGAN-HD · UCF-101 | FVD: 838.09 -> 700.00 | Helped |
| Add the contrastive loss (ablation) | MoCoGAN-HD · (FFHQ, VoxCeleb) | ACD: 0.5328 -> 0.4353 | Helped |
| Add the contrastive loss (ablation) | MoCoGAN-HD · (FFHQ, VoxCeleb) | Human Preference (%): 31.7 -> 68.3 | Helped |
| Add the mutual information loss (ablation) | MoCoGAN-HD · (FFHQ, VoxCeleb) | ACD: 0.5158 -> 0.4353 | Helped |
| Add the mutual information loss (ablation) | MoCoGAN-HD · (FFHQ, VoxCeleb) | Human Preference (%) for motion diversity: 35.6 -> 64.4 | Helped |
| Use a fixed pretrained image generator and learn video motion as trajectories in its latent space | MoCoGAN-HD · UCF-101 | IS: 26.60 -> 33.95 | Helped |
| Use a fixed pretrained image generator and learn video motion as trajectories in its latent space | MoCoGAN-HD · FaceForensics | ACD: 0.4914 -> 0.3300 | Helped |
| Use a fixed pretrained image generator and learn video motion as trajectories in its latent space | MoCoGAN-HD · Sky Time-lapse | FVD: 451.14 -> 77.77 | Helped |
Closest papers by meaning
- Decouple Content and Motion for Conditional Image-to-Video Generation (AAAI 2024 · similarity 0.88)
- Snap Video: Scaled Spatiotemporal Transformers for Text-to-Video Synthesis (CVPR 2024 · similarity 0.88)
- Wan-Move: Motion-controllable Video Generation via Latent Trajectory Guidance (NeurIPS 2025 · similarity 0.88)
- Make Pixels Dance: High-Dynamic Video Generation (CVPR 2024 · similarity 0.88)
- Frame In-N-Out: Unbounded Controllable Image-to-Video Generation (NeurIPS 2025 · similarity 0.87)
- Generative Rendering: Controllable 4D-Guided Video Generation with 2D Diffusion Models (CVPR 2024 · 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.