GraphVid controls video generation with interaction graphs and cuts FID by 39.9%
PLAN-Lab · hf · 2026-07-24
What it is
GraphVid is a graph-conditioned image-to-video model that lets users control multi-object interactions through structured interaction graphs instead of brittle text prompts or manual motion tracks.
Why it matters
- Trajectory-based control does not scale well when scenes get crowded or objects overlap.
- GraphVid uses a semantic interface built around interaction graphs, making control more flexible and more precise for multi-subject scenes.
- The authors also release GraphVid-Bench, a large-scale interaction-centric dataset with structured relational annotations.
Results
- Despite using less training data and fewer trainable parameters than prior motion-control methods, GraphVid reports strong controllability and video quality.
- Compared with Motion-I2V, it claims up to 39.9% lower FID and 37.6% lower FVD, plus better PSNR (9.87 → 15.98) and SSIM (0.38 → 0.61).
- The paper argues that structured semantic interfaces are a promising direction for controllable video generation.
More from Multimodal
- ostris pre-training "Primal VAE": 1-channel latents match modern 8x VAEs via patching — ostrisai · 2026-10-26
- Slopcore: AI-generated meme films turn low-fi slop aesthetics into a subculture — Yamapama · 2026-10-26
- TischEins, a large Midjourney SREF library, relaunches with new design after security outage — tisch_eins · 2026-10-26
- MOSS-Image-SFT 4-step distillation tops 16.5K downloads on Hugging Face — _akhaliq · 2026-10-26
- Claude Code can now edit videos: a 10-prompt workflow from raw footage to final cut — TawohAwa · 2026-10-26
- One-sentence 3D worlds turn cinematic with CineMap and TapNow — gizakdag · 2026-10-26