Chaosaiart-Nodes

★ 115

低显存逐帧动画txt2video/img2video/video2videoPrompt/Checkpoint/缓存

低显存GPU优化的逐帧动画节点，支持txt2video、img2video、video2video，含Prompt/Checkpoint切换、缓存、按帧计数与Ksampler兼容，便于在低显存下生成动画。

💡 在低显存显卡上逐帧生成文本或图像到视频的动画。

🍴 13 Forks💻 Python🔄 2025-05-06

🔗 GitHub 原文

📦

网盘下载

复制链接后前往夸克网盘下载

https://pan.quark.cn/s/b6135d9bd930

📦 requirements.txt

opencv-python
tqdm

📄 README

LowVRAM Animation

Work in Progress.

Simplified workflow, VAE inclusion in KSampler

Workflow automation, e.g., changing checkpoints

Logical process for testing elements like Controlnet

Rethinking workflows with switches

This node is based on frame-to-frame generation (img after img).

This means it is VRAM efficient and suitable for GPUs with low VRAM.

Additional: Chaosaiart-Node is in the early phase,
more nodes will be added + Bug fixes + changes.

Install ComfyUI

Download Chaosaiart-Nodes

Unzip Chaosaiart-Nodes with 7zip

Grab the Unzipped “Chaosaiart-Nodes” folder

go into ComfyUI Folder, then “../comfyUI/custom_nodes/” place it in.

Windows:

Use the Install_windows file in the “Chaosaiart-Nodes” folder

Linux:

pip install opencv-python & pip install tqdm

If you are using a virtual environment (venv), make sure it is activated before installation.

Open ComfyUI, Load or Drag & Drop one of the workflows.

Install ComfyUI

Install ComfyUI Manager

Open Manager, Use “Install via Git URL”.

Command:

https://github.com/chaosaiart/Chaosaiart-Nodes

Open ComfyUI, Load or Drag & Drop one of the workflows.

Change Checkpoint, you can specify when.

1x Frame = 1x generated Img / Batch

Change Prompt, you can specify when.

1x Frame = 1x generated Img / Batch

Change Checkpoint & Prompt, you can specify when.

1x Frame = 1x generated Img / Batch

Basic img2img Batch Loader

Select folder, Each generation uses the next image.

You also can Repeat Img.

Basic Controlnet Batch Loader

Select folder, Each generation uses the next image.

You also can Repeat Img.

Splitt Video in to Frames / img

Stitching Image to Video

Stitching Image to Gif

First Step for Animation

Cache a Img for the next generate, in this Case Img2img.

My Promptchanger workflow has been implemented. Prompt Changer

Change Controlnet Settings by Counting.

1x Frame = 1x generated Img / Batch = Step

You will need Canny Model

Save and Load your Prompt + Simple Workflow.

For Loading & Saving Text you need to Install this Custom Node

Using Any-Switch to control any input at any time.

Utilize arrays to enhance the organization of your workspace, or swiftly switch between processes to test things.

idrirap has written a great Lora Auto-Trigger-Words Custom Node. Through this, I found out that all tags are included in the metadata. From now on, all tags and their frequency of use will be output by Lora via Info.

How to Use Loras:

usable in all 🔶Chaosaiart prompt fields.

Change Contrast, Color, Brightness, Red, Green, Blue.

Animation:

This workflow has integrated the cache of the Cache Workflow in KSampler.

Additionally, KSampler has been modified to yield better results for animations.

For example, the method “Fixed 0.5” has been optimized for checkpoint sd1.5.

Afterwards, you must stitch the images into a video using img2video Basic Workflow

Same workflow as txt2video, but with image input.

In this workflow, the 🔶 Ksampler txt & img2video v1 has been dissected into its individual components.
This allows you to see what it consists of and enables you to adjust the values yourselves.