# I3DM: Implicit 3D-aware Memory Retrieval and Injection for Consistent Video Scene Generation

[**Project page**](https://riga2.github.io/i3dm/) <be> & [**Paper**](https://arxiv.org/abs/2603.23413)

Weights:
- Retrieval Net: ckpt_0000000000016000.pt
- LVSM Decoder Model (Finetuned on 4-frame inputs): LVSM_F4.pt
- Fined Video Diffusion Model: step-11000.safetensors