Accepted to CVPR 2025!

Ryan Burgert^1,3, Yuancheng Xu^1,4, Wenqi Xian¹, Oliver Pilarski¹, Pascal Clausen¹, Mingming He¹, Li Ma¹,

Yitong Deng^2,5, Lingxiao Li², Mohsen Mousavi¹, Michael Ryoo³, Paul Debevec¹, Ning Yu^1†

¹Netflix Eyeline Studios, ²Netflix, ³Stony Brook University, ⁴University of Maryland, ⁵Stanford University
^†Project Lead

Table of Contents

• Abstract
• Quick Start: Cut-and-drag Motion Control
  • Animation Template GUI (Local)
  • Running Video Diffusion (GPU)
• TODO
• Citation

📖 Abstract

Go-with-the-Flow is an easy and efficient way to control the motion patterns of video diffusion models. It lets a user decide how the camera and objects in a scene will move, and can even let you transfer motion patterns from one video to another.

We simply fine-tune a base model — requiring no changes to the original pipeline or architecture, except: instead of using pure i.i.d. Gaussian noise, we use warped noise instead. Inference has exactly the same computational cost as running the base model.

If you create something cool with our model - and want to share it on our website - email [email protected]. We will be creating a user-generated content section, starting with whomever submits the first video!

If you like this project, please give it a ★!

🔥 Community Adoption

A huge thank you to all who contributed - videos to be added here soon!

• ComfyUI implementation by @kijai
• HuggingFace Space #1 by fffiloni
• HuggingFace Space #2 by Ryan Burgert
• AnimateDiff Implementation by spacepxl
• HunyuanVideo Implementation by spacepxl
• Cut-and-drag using SAMv2 and its web interface by Pablo Salamanca
• Japanese Tutorial by Takamasa Tamura

Examples:

🚀 Quick Start: Cut-and-drag Motion Control

Cut-and-drag motion control lets you take an image, and create a video by cutting out different parts of that image and dragging them around.

For cut-and-drag motion control, there are two parts: an GUI to create a crude animation (no GPU needed), then a diffusion script to turn that crude animation into a pretty one (requires GPU).

YouTube Tutorial:

1. Animation Template GUI (Local)

1. Clone this repo, then cd into it.

2. Install local requirements:

   pip install -r requirements_local.txt

3. Run the GUI:

   python cut_and_drag_gui.py

4. Follow the instructions shown in the GUI.

After completion, an MP4 file will be generated. You'll need to move this file to a computer with a decent GPU to continue.

2. Running Video Diffusion (GPU)

1. Clone this repo on the machine with the GPU, then cd into it.

2. Install requirements:

   pip install -r requirements.txt

3. Warp the noise (replace <PATH TO VIDEO OR URL> accordingly):

   python make_warped_noise.py <PATH TO VIDEO OR URL> --output_folder noise_warp_output_folder

4. Run inference:

   python cut_and_drag_inference.py noise_warp_output_folder \
       --prompt "A duck splashing" \
       --output_mp4_path "output.mp4" \
       --device "cuda" \
       --num_inference_steps 30
   

Adjust folder paths, prompts, and other hyperparameters as needed. The output will be saved as output.mp4.

📋 TODO

• Upload All CogVideoX Models
• Upload Cut-And-Drag Inference Code
• Release to Arxiv
• Google Colab for people without GPU's
• Depth-Warping Inference Code
• T2V Motion Transfer Code
• I2V Motion Transfer Code (allows for first-frame editing)
• ComfyUI Node
• Release 3D-to-Video Inference Code + Blender File
• Upload AnimateDiff Model
• Replicate Instance
• Fine-Tuning Code

✒️ Citation

If you use this in your research, please consider citing:

@inproceedings{burgert2025gowiththeflow,
  title={Go-with-the-Flow: Motion-Controllable Video Diffusion Models Using Real-Time Warped Noise},
  author={Burgert, Ryan and Xu, Yuancheng and Xian, Wenqi and Pilarski, Oliver and Clausen, Pascal and He, Mingming and Ma, Li and Deng, Yitong and Li, Lingxiao and Mousavi, Mohsen and Ryoo, Michael and Debevec, Paul and Yu, Ning},
  booktitle={CVPR},
  year={2025}
}