---
license: apache-2.0
pipeline_tag: image-to-video
---

# LongE2V

Released LoRA weights for **LongE2V**, fine-tuned on top of `THUDM/CogVideoX-5b-I2V`.

This model is presented in the paper [LongE2V: Long-Horizon Event-based Video Reconstruction, Prediction, and Frame Interpolation with Video Diffusion Models](https://huggingface.co/papers/2607.08770).

- **Project Page**: https://cdfan0627.github.io/LongE2V-page/
- **Code**: [cdfan0627/LongE2V](https://github.com/cdfan0627/LongE2V) (contains training and evaluation scripts for this model)
- **Data**: preprocessed evaluation/training data (ECD, HQF, MVSEC) is released at [fansam39/LongE2V-data](https://huggingface.co/datasets/fansam39/LongE2V-data).

## Files

- `pytorch_lora_weights.safetensors`: LoRA weights for the CogVideoX-5b-I2V base model.

## Usage

```bash
WEIGHT_URL=https://huggingface.co/fansam39/LongE2V/resolve/main/pytorch_lora_weights.safetensors bash scripts/download_weights.sh
```

Or with the Hugging Face CLI:

```bash
hf download fansam39/LongE2V pytorch_lora_weights.safetensors --local-dir weights
```

See the code repository's README for full inference and evaluation instructions.