smthemex
caf2f1a275
Updated README to reflect changes in VRAM requirements and added new links for fp8 and upscale lora models.
62 lines
2.3 KiB
Markdown
62 lines
2.3 KiB
Markdown
# ComfyUI_ChronoEdit_SM
|
||
[ChronoEdit](https://github.com/nv-tlabs/ChronoEdit): Towards Temporal Reasoning for Image Editing and World Simulation,you can use this node in comfyUI,and Vram >8G
|
||
|
||
# Update
|
||
* support upscale lora and fp8 dit，if use 8 step and upscale need step>12,if use upscale only need steps>35
|
||
* 支持fp8量化模型和放大lora，两个一起用时最好步数大一点，不用加速时，步数大于35
|
||
* 推荐5帧8步lora/ use 5 frames 8 steps lora
|
||
* 如果跑29帧要跑50步/ if use 29 frames need 50 steps
|
||
|
||
|
||
1.Installation
|
||
-----
|
||
In the ./ComfyUI/custom_nodes directory, run the following:
|
||
```
|
||
git clone https://github.com/smthemex/ComfyUI_ChronoEdit_SM
|
||
|
||
```
|
||
|
||
2.requirements
|
||
----
|
||
|
||
```
|
||
pip install -r requirements.txt
|
||
```
|
||
|
||
3.checkpoints
|
||
----
|
||
|
||
1. gguf [links](https://huggingface.co/QuantStack/ChronoEdit-14B-GGUF/tree/main)
|
||
2. dit fp8 [links](https://huggingface.co/cocorang/ChronoEdit-14B-Diffusers-FP8) optional
|
||
3. 8 steps lora [links](https://huggingface.co/nvidia/ChronoEdit-14B-Diffusers/tree/main/lora)
|
||
4. upscale lora [links](https://huggingface.co/nvidia/ChronoEdit-14B-Diffusers-Upscaler-Lora)
|
||
5 .wan T5 clipvison vae [ links](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged)
|
||
```
|
||
├── ComfyUI/models/gguf # or fp8
|
||
| ├── ChronoEdit-14B-Q6_K.gguf # or Q8
|
||
├── ComfyUI/models/vae
|
||
| ├──Wan2.1_VAE.pth
|
||
├── ComfyUI/models/clip
|
||
| ├──umt5_xxl_fp8_e4m3fn_scaled.safetensors
|
||
| ├──qwen_2.5_vl_7b.safetensors # OPTIONAL if use vl 可选 太慢
|
||
├── ComfyUI/models/clip_vision
|
||
| ├──clip_vision_h.safetensors
|
||
├── ComfyUI/models/loras
|
||
| ├──chronoedit_distill_lora.safetensors
|
||
| ├──upsample_lora_diffusers.safetensors
|
||
|
||
```
|
||
|
||
# 4 Example
|
||
|
||
|
||
# 5 Citation
|
||
```
|
||
@article{wu2025chronoedit,
|
||
title={ChronoEdit: Towards Temporal Reasoning for Image Editing and World Simulation},
|
||
author={Wu, Jay Zhangjie and Ren, Xuanchi and Shen, Tianchang and Cao, Tianshi and He, Kai and Lu, Yifan and Gao, Ruiyuan and Xie, Enze and Lan, Shiyi and Alvarez, Jose M. and Gao, Jun and Fidler, Sanja and Wang, Zian and Ling, Huan},
|
||
journal={arXiv preprint arXiv:2510.04290},
|
||
year={2025}
|
||
}
|
||
```
|