🎬 LongE2V: Event-based Video Reconstruction

Reconstruct grayscale video with a fine-tuned CogVideoX-5B-I2V diffusion model. Upload an ordinary video clip — the demo synthesizes the event stream on the backend (a contrast-threshold / ESIM-style event simulator) and voxelizes it into the event-voxel format the model expects. Advanced users can still upload raw event .npz voxel grids directly.

Paper | Code | Weights