UPDATE

Added "cond-image-leakage" (CIL) versions of the 1024 and 512 models from https://huggingface.co/GraceZhao/DynamiCrafter-CIL-1024

https://github.com/thu-ml/cond-image-leakage

---

For reference on model card metadata, see the spec: https://github.com/huggingface/hub-docs/blob/main/modelcard.md?plain=1

Doc / guide: https://huggingface.co/docs/hub/model-cards

{}

Bf16 safetensors versions of the DynamiCrafter models by Doubiiu: https://huggingface.co/Doubiiu

---

This is a video diffusion model that takes in a single or two still images as a conditioning
image and text prompt describing dynamics, and generates looping videos or interpolation from them.

Model Details

Model Description

DynamiCrafter, a (Text-)Image-to-Video/Image Animation approach, aims to generate
short video clips (~2 seconds) from a conditioning image and text prompt.

This model was trained to generate 16 video frames at a resolution of 320x512
given a context frame of the same resolution.

• Developed by: CUHK & Tencent AI Lab
• Funded by: CUHK & Tencent AI Lab
• Model type: Generative frame interpolation and looping video generation
• Finetuned from model: VideoCrafter1 (320x512)

Model Sources

For research purpose, we recommend our Github repository (https://github.com/Doubiiu/DynamiCrafter),
which includes the detailed implementations.

• Repository: https://github.com/Doubiiu/DynamiCrafter
• Paper: https://arxiv.org/abs/2310.12190

Uses

Direct Use

We develop this repository for RESEARCH purposes, so it can only be used for personal/research/non-commercial purposes.

Limitations

• The generated videos are relatively short (2 seconds, FPS=8).
• The model cannot render legible text.
• Faces and people in general may not be generated properly.
• The autoencoding part of the model is lossy, resulting in slight flickering artifacts.

How to Get Started with the Model

Check out https://github.com/Doubiiu/DynamiCrafter