Changed lots of things to better integrate this to ComfyUI, you can (and have to) use clip_vision and clip models, but memory usage is much better and I was able to do 512x320 under 10GB VRAM. New example workflows are included, all old workflows will have to be updated.
Initial ToonCrafter support with it's own node.
VRAM required for 512x320 seems to be slightly under 15GB for sampling, and peaks to ~17GB at decoding. Dropping the resolution a bit can reduce the memory use a lot.
Decoding is the most memory hungry operation in all this, and currently REQUIRES XFORMERS for full quality results, however it is possible to use the standard Comfy VAE decoder for bit less quality but far less memory used. The ToonCrafter model can also be used with the old Dynamicrafter I2V -node, quality then suffers a lot more, memory usage is halved, fitting under 8GB at best.
Fp8 option can also further reduce memory use by 1-2GB.
chrome_TPISlAxki3.mp4
Either manager and install from git, or clone this repo to custom_nodes and run:
pip install -r requirements.txt
or if you use portable (run this in ComfyUI_windows_portable -folder):
python_embeded\python.exe -m pip install -r ComfyUI\custom_nodes\ComfyUI-DynamiCrafterWrapper\requirements.txt
Currently even if this can run without xformers, the memory usage is huge. Recommended to use xformers if possible:
pip install xformers --no-deps
or with portable:
python_embeded\python.exe -m pip install xformers --no-deps
UPDATE: Converted the models to bf16 and .safetensors format here: https://huggingface.co/Kijai/DynamiCrafter_pruned/tree/main
Models go to ComfyUI/models/checkpoints (can also be in subfolder, up to you)
If you want to use the original models, they are available here, they do need to be renamed to be used with the node:
Name this: dynamicrafter_1024_v1.ckpt
https://huggingface.co/Doubiiu/DynamiCrafter_1024
Interpolation model should be named: dynamicrafter_512_interp_v1.ckpt
https://huggingface.co/Doubiiu/DynamiCrafter_512_Interp/
With fp16 1024x576 uses bit under 10GB VRAM, and interpolation at 512p can be done with 8GB
Looping example:
chrome_NxIN4fASYw.mp4
Interpolation example:
chrome_dIEgzBhdua.mp4
chrome_KHldmHYu1E.mp4
https://github.com/Doubiiu/DynamiCrafter
Jinbo Xing, Menghan Xia*, Yong Zhang, Haoxin Chen, Wangbo Yu,
Hanyuan Liu, Xintao Wang, Tien-Tsin Wong*, Ying Shan
(* corresponding authors)
From CUHK and Tencent AI Lab.
🔥🔥 Generative frame interpolation / looping video generation model weights (320x512) have been released!
🔥 New Update Rolls Out for DynamiCrafter! Better Dynamic, Higher Resolution, and Stronger Coherence!
🤗 DynamiCrafter can animate open-domain still images based on text prompt by leveraging the pre-trained video diffusion priors. Please check our project page and paper for more information.
😀 We will continue to improve the model's performance.
👀 Seeking comparisons with Stable Video Diffusion and PikaLabs? Click the image below.
|
|
|
|
|
|
|
|
| "bear playing guitar happily, snowing" | "boy walking on the street" | ||
|
|
|
|
 |
| Input starting frame | Input ending frame | Generated video |
|
|
|
|
|
|
|
|
|
|
|
|
- [2024.03.14]: 🔥🔥 Release generative frame interpolation and looping video models (320x512).
- [2024.02.05]: Release high-resolution models (320x512 & 576x1024).
- [2023.12.02]: Launch the local Gradio demo.
- [2023.11.29]: Release the main model at a resolution of 256x256.
- [2023.11.27]: Launch the project page and update the arXiv preprint.
| Model | Resolution | GPU Mem. & Inference Time (A100, ddim 50steps) | Checkpoint |
|---|---|---|---|
| DynamiCrafter1024 | 576x1024 | 18.3GB & 75s (perframe_ae=True) |
Hugging Face |
| DynamiCrafter512 | 320x512 | 12.8GB & 20s (perframe_ae=True) |
Hugging Face |
| DynamiCrafter256 | 256x256 | 11.9GB & 10s (perframe_ae=False) |
Hugging Face |
| DynamiCrafter512_interp | 320x512 | 12.8GB & 20s (perframe_ae=True) |
Hugging Face |
Currently, our DynamiCrafter can support generating videos of up to 16 frames with a resolution of 576x1024. The inference time can be reduced by using fewer DDIM steps.
GPU memory consumed on RTX 4090 reported by @noguchis in Twitter: 18.3GB (576x1024), 12.8GB (320x512), 11.9GB (256x256).
conda create -n dynamicrafter python=3.8.5
conda activate dynamicrafter
pip install -r requirements.txt- Download pretrained models via Hugging Face, and put the
model.ckptwith the required resolution incheckpoints/dynamicrafter_[1024|512|256]_v1/model.ckpt. - Run the commands based on your devices and needs in terminal.
# Run on a single GPU:
# Select the model based on required resolutions: i.e., 1024|512|320:
sh scripts/run.sh 1024
# Run on multiple GPUs for parallel inference:
sh scripts/run_mp.sh 1024Download pretrained model DynamiCrafter512_interp and put the model.ckpt in checkpoints/dynamicrafter_512_interp_v1/model.ckpt.
sh scripts/run_application.sh interp # Generate frame interpolation
sh scripts/run_application.sh loop # Looping video generation- Download the pretrained models and put them in the corresponding directory according to the previous guidelines.
- Input the following commands in terminal (choose a model based on the required resolution: 1024, 512 or 256).
python gradio_app.py --res 1024Download the pretrained model and put it in the corresponding directory according to the previous guidelines.
python gradio_app_interp_and_loop.py Community Extensions for Image-to-Video: ComfyUI (Thanks to chaojie).
VideoCrafter1: Framework for high-quality video generation.
ScaleCrafter: Tuning-free method for high-resolution image/video generation.
TaleCrafter: An interactive story visualization tool that supports multiple characters.
LongerCrafter: Tuning-free method for longer high-quality video generation.
MakeYourVideo, might be a Crafter:): Video generation/editing with textual and structural guidance.
StyleCrafter: Stylized-image-guided text-to-image and text-to-video generation.
@article{xing2023dynamicrafter,
title={DynamiCrafter: Animating Open-domain Images with Video Diffusion Priors},
author={Xing, Jinbo and Xia, Menghan and Zhang, Yong and Chen, Haoxin and Yu, Wangbo and Liu, Hanyuan and Wang, Xintao and Wong, Tien-Tsin and Shan, Ying},
journal={arXiv preprint arXiv:2310.12190},
year={2023}
}We would like to thank AK(@_akhaliq) for the help of setting up hugging face online demo, and camenduru for providing the replicate & colab online demo.
We develop this repository for RESEARCH purposes, so it can only be used for personal/research/non-commercial purposes.