DreamRunner: Fine-Grained Storytelling Video Generation with Retrieval-Augmented Motion Adaptation

Zun Wang, Jialu Li, Han Lin, Jaehong Yoon, Mohit Bansal

Code coming soon!

ToDos

• Release the inference code on T2V-ComBench.
• Release the code for retrieving videos and training character and motion loras.
• Release the inference code for storytelling video genetation.

Setup

Environment Setup

conda create -n dreamrunner python==3.10
conda activate dreamrunner
pip install -r requirements.txt 

Download Models

DreamRunner is implemented using CogVideoX-2B. You can download it here and put it to pretrained_models/CogVideoX-2b.

Running the Code

T2V-Combench

Inference

We provide the plans we used for T2V-ComBench in MotionDirector_SR3AI/t2v-combench/plan. You can specify the GPUs you want use in MotionDirector_SR3AI/t2v-combench-2b.sh for parallel inference. Then directly Infer 600 videos on 6 dimensions of T2V-ComBnech with the following script

cd MotionDirector_SR3AI
bash run_bench_2b.sh

The generated videos will be saved at MotionDirector_SR3AI/T2V-CompBench.

Evaluation

Please follow T2V-ComBench for evaluating the generated videos.

Storytell Video Generation

Coming soon!

Citation

If you find our project useful in your research, please cite the following paper:

@article{zun2024dreamrunner,
    author = {Zun Wang and Jialu Li and Han Lin and Jaehong Yoon and Mohit Bansal},
    title  = {DreamRunner: Fine-Grained Storytelling Video Generation with Retrieval-Augmented Motion Adaptation},
	journal   = {arxiv},
	year      = {2024},
	url       = {https://arxiv.org/abs/2411.16657}
}