ComfyUI-easygoing-nodes v2.0: Better Workflow Screenshots, Prompt Embedding & VRAM Control

Introduction

Hello, this is Kimama / Easygoing. Today, I’d like to introduce the v2.0.0 update of the ComfyUI custom node pack ComfyUI-easygoing-nodes.

What is ComfyUI-easygoing-nodes?

ComfyUI-easygoing-nodes is a custom node pack for ComfyUI that I released in November 2025.

Initial release of ComfyUI-easygoing-nodes

ComfyUI-easygoing-nodes includes features I use in my daily workflow, such as saving screenshots of workflows and embedding prompt metadata when saving images.

This time, I carried out a major update to support ComfyUI’s Nodes 2.0 and V3 schema.

About ComfyUI’s Nodes 2.0 and V3 schema

I’d like to reintroduce the features of ComfyUI-easygoing-nodes, including the newly added ones.

1. Workflow Screenshots (Nodes 1.0)

ComfyUI-easygoing-nodes includes a feature that takes a screenshot of the ComfyUI workflow with embedded metadata.

Right-click → Workflow image → Export PNG

This feature was originally included in comfyui-custom-scripts, but repeated updates to ComfyUI_frontend caused display issues.

By including an improved version of the code in ComfyUI-easygoing-nodes, screenshots can now be taken correctly in most cases.

With the widespread adoption of high-resolution devices such as 4K monitors and iPads, this feature saves screenshots at 4 times the original resolution so that they display cleanly even on high-resolution screens.

Note that the new ComfyUI UI currently in beta, Nodes 2.0, uses a complex combination of vue.js and litegraph.js for rendering, so workflow export for Nodes 2.0 is not supported at this time.

When using the workflow-saving feature of ComfyUI-easygoing-nodes, please switch to Nodes 1.0.

2. Improved Model Loading

ComfyUI-easygoing-nodes includes the Load Checkpoint (with name clean) and Load Diffusion Model (with name clean) nodes for loading models.

Both of these nodes display a list of models stored in both of the following folders:

• ComfyUI/models/checkpoints
• ComfyUI/models/diffusion_models

Additionally, the model_name_clean output removes the model’s file extension and format indicators.

Strings removed (including uppercase variants)

fp32, fp16, bf16, fp8,
svd, scaled, mxfp8, nvfp4, int8, convrot, hq,
e4m3fn, e4m3, e5m2fn, e5m2,
.safetensors, .pth, .pt
trailing “-” and “_”

In ComfyUI, adding a slash to the file_path of a Save Image node allows you to save images into arbitrary subfolders.

By appending a slash after the model name in this way, you can organize and manage output images by model.

3. Freeing VRAM and RAM After CLIP Processing

ComfyUI tries to keep models loaded in VRAM as much as possible, and when VRAM runs short, it moves older models to RAM.

Models moved to RAM are kept there (pinned_memory) if they are judged likely to be reused, to avoid reloading from storage.

This model retention works as intended when there is plenty of VRAM and RAM, but on systems with limited memory it can reduce available working memory and actually decrease stability and performance.

Using the CLIP Text Encode (with offload) node in place of the standard Clip Text Encode node explicitly frees the text encoder from VRAM and RAM after processing, freeing up memory for other operations.

Depending on your environment, CLIP Text Encode (with offload) may improve or decrease performance, so please test whether it suits your setup.

4. Saving Images with Prompts Attached

When you save an image in ComfyUI, the workflow is saved together as JSON metadata. However, this data is complex, and finding the prompts can often be difficult.

Excerpt of ComfyUI workflow metadata

{"6": {"inputs": {"mask": ["195", 0]}, "class_type": "MaskPreview", "_meta": {"title": "Preview Mask"}}, "7": {"inputs": {"width": 1152, "height": 832, "batch_size": 1}, "class_type": "EmptyLatentImage", "_meta": {"title": "Empty Latent Image"}}, "8": {"inputs": {"noise_seed": ["41", 0]}, "class_type": "RandomNoise", "_meta": {"title": "RandomNoise"}}, "9": {"inputs": {"scheduler": "beta", "steps": 15, "denoise": 1.0, "model": ["232", 0]}, "class_type": "BasicScheduler", "_meta": {"title": "BasicScheduler"}}, "10": {"inputs": {"clip_name": "CLIP-SAE-ViT-L-14-FP32.safetensors"}, "class_type": "CLIPVisionLoader", "_meta": {"title": "Load CLIP Vision"}}, "11": {"inputs": {"mask": ["163", 0]}, "class_type": "MaskPreview", "_meta": {"title": "Preview Mask"}}, "14": {"inputs": {"model_name": "4x_IllustrationJaNai_V1_ESRGAN_135k.safetensors"}, "class_type": "UpscaleModelLoader", "_meta": {"title": "Load Upscale Model"}},

The Save Image with Prompt node therefore saves the following five items as separate metadata in addition to the ComfyUI workflow:

• positive_prompt
• additional_prompt
• negative_prompt
• caption
• seed

Note that in ComfyUI, seeds are normally handled as Int (integer) values, but inserting a Preview as text node converts them into the string format usable by the Save Image with Prompt node.

XnView MP metadata display

digiKam metadata display

When you view illustrations generated in ComfyUI with a regular image viewer, the metadata is displayed at a glance, making it easy to copy prompts or organize images.

5. Color Adjustment and HDR Processing

The HDR effect with LAB Adjusts node allows color adjustments in the LAB color space.

About the LAB color space

Unlike the usual RGB color space, LAB separates brightness and color, making unintended changes less likely.

HDR Effects with LAB Adjusts node

Comparison image

The HDR Effects with LAB Adjusts node is well suited for producing the high-contrast illustrations that are particularly popular in Western markets.

6. Detecting Differences Between Images

The Image Difference Checker node takes two similar images as input and performs the following analysis:

• Creation of a difference map
• Numerical similarity evaluation using MAE and SSIM
• Color analysis with tone curves

VAE accuracy comparison using the Image Difference Checker node

When you change the settings of Image Difference Checker and want to re-output the results, select the node and click the blue partial execution button at the top. This re-runs only the nodes from Image Difference Checker onward.

The sdxl_clip.py Replacement Feature Has Been Removed

ComfyUI-easygoing-nodes previously included a feature that improved SDXL image quality by replacing sdxl_clip.py. However, recent ComfyUI updates rendered it ineffective, so the sdxl_clip.py replacement feature has been removed in this update.

About the sdxl_clip.py replacement feature

How to Install ComfyUI-easygoing-nodes

Finally, here’s how to install ComfyUI-easygoing-nodes.

ComfyUI-easygoing-nodes is registered in the ComfyUI Registry, so you can install it using ComfyUI’s standard Nodes Manager.

Important Notes for Updating from v1.8.5 to v2.0.0!

ComfyUI-easygoing-nodes underwent a major update from v1.8.5 to v2.0.0 to support ComfyUI’s Nodes 2.0 and V3 schema.

When you update ComfyUI-easygoing-nodes to v2.0.0 or later and open an older workflow, the nodes are designed to automatically migrate to the v2.0.0 versions. However, automatic migration may not work correctly depending on your environment.

In such cases, please delete the old nodes and place the new ones again.

Summary: Give ComfyUI-easygoing-nodes a Try

• Update ComfyUI-easygoing-nodes to v2.0.0
• Support for Nodes 2.0 and V3 schema
• You may need to delete and re-place older nodes

This article introduced the update to ComfyUI-easygoing-nodes.

ComfyUI’s core updates extremely quickly and often includes breaking changes without mercy, so keeping custom nodes working requires considerable effort.

I’m still a programming beginner, so I rely almost entirely on AI (mainly Claude) to create custom nodes. Lately, the mixture of old and new ComfyUI specifications has led to more cases where AI gets confused, creating a new kind of difficulty compared to before.

Nevertheless, the AI used for coding is clearly becoming smarter as well, so I plan to continue maintaining the pack with AI’s help going forward.

Thank you for reading all the way to the end!