license: apache-2.0
pipeline_tag: image-to-video
library_name: diffusers
tags:
- video-outpainting
- diffusion
- video-generation
M3DDM+: An Improved Video Outpainting by a Modified Masking Strategy
Takuya Murakawa, Takumi Fukuzawa, Ning Ding, Toru Tamaki
Nagoya Institute of Technology
IWAIT 2026
M3DDM+ provides a computationally efficient framework for video outpainting via latent diffusion modeling. It addresses significant quality degradation—such as spatial blur and temporal inconsistency—found in previous methods like M3DDM, especially under challenging scenarios with limited camera motion or large outpainting regions. By applying a uniform mask direction and width across all frames during training, M3DDM+ substantially improves visual fidelity and temporal coherence while maintaining computational efficiency.
Sample Usage
For detailed environment setup and to download necessary pre-trained models, please refer to the GitHub repository. Once set up, you can run inference using the following command:
CUDA_VISIBLE_DEVICES=0 python src/inference.py \
--input_video_path "path/to/input_video.mp4" \
--pretrained_sd_dir "stable-diffusion-v1-5" \
--video_outpainting_model_dir "MurakawaTakuya/M3DDM-Plus" \
--output_dir "path/to/output_directory" \
--target_ratio_list "1:1" \
--output_size 256
Citation
If our work is helpful, please help to ⭐ the repo.
Please consider citing our paper if you found our work interesting and useful.
@article{murakawa2026m3ddmplus,
title={M3DDM+: An improved video outpainting by a modified masking strategy},
author={Murakawa, Takuya and Fukuzawa, Takumi and Ding, Ning and Tamaki, Toru},
journal={International Workshop on Advanced Image Technology (IWAIT)},
year={2026}
}
Acknowledgement
The inference and pipeline code is based on published code of M3DDM-Video-Outpainting. The training and evaluation code was reproduced based on the M3DDM paper as it isn't published, and modified for our proposed method.