Save 4 hours! We auto-setup your workflow! Free!

Drop your workflow.json — we handle every dependency, custom node, and model. Just open the link and run.

Auto-Setup Workflow Json (Free) Now!
ComfyUI > Nodes > ComfyUI-ARZUMATA

ComfyUI Extension: ComfyUI-ARZUMATA

Repo Name

ComfyUI-ARZUMATA

Author
arzumata (Account age: 2640 days)
Nodes
View all nodes(1)
Latest Updated
2026-05-21
Github Stars
0.01K

How to Install ComfyUI-ARZUMATA

Install this extension via the ComfyUI Manager by searching for ComfyUI-ARZUMATA
  • 1. Click the Manager button in the main menu
  • 2. Select Custom Nodes Manager button
  • 3. Enter ComfyUI-ARZUMATA in the search bar
After installation, click the Restart button to restart ComfyUI. Then, manually refresh your browser to clear the cache and access the updated list of nodes.

Visit ComfyUI Online for ready-to-use ComfyUI environment

  • Free trial available
  • 16GB VRAM to 80GB VRAM GPU machines
  • 400+ preloaded models/nodes
  • Freedom to upload custom models/nodes
  • 200+ ready-to-run workflows
  • 100% private workspace with up to 200GB storage
  • Dedicated Support

Run ComfyUI Online

ComfyUI-ARZUMATA Description

ComfyUI-ARZUMATA adds nodes to select model devices, cache CLIP prompt encoding for repeated runs, and load images by file path, including merged Krita projects with masks for image workflows.

ComfyUI-ARZUMATA Introduction

ComfyUI-ARZUMATA is a collection of custom nodes for ComfyUI. Its nodes address three practical workflow needs: choosing where a model runs, avoiding repeated text-encoding work, and loading images directly from file paths.

You might use it when you want to try running a model on a particular device, repeatedly generate images from the same prompt, or bring an image from a Krita project into a workflow. The nodes are independent tools, so you can use the ones that fit your workflow without using the others.

How ComfyUI-ARZUMATA Works

A ComfyUI workflow passes information between connected nodes. ComfyUI-ARZUMATA adds nodes that change how particular information is handled:

  • Device override: Directs a connected model or module to a device you specify, such as cpu or cuda:0.
  • Text-encoding cache: Remembers the result of converting prompt text into conditioning, the information a generation workflow uses to interpret your prompt. When the same inputs are used again, the node can reuse the earlier result instead of repeating that work.
  • Image loading: Reads an image from a file path and provides it to the workflow as an image, a mask, and the image path.

Think of the cache as keeping a prepared version of a prompt close at hand. If you generate several variations with the same text, the node can reuse that preparation. If you change the text or the CLIP model, it prepares a new result. In the FLUX caching node, changing the guidance value alone does not invalidate the cached result.

ComfyUI-ARZUMATA Features

Universal Device Override

Universal Device Override lets you specify the device on which a model or module should run. It is designed to work with MODEL, CLIP, and VAE, as well as types such as WANVAE, CLIP_VISION, CONTROL_NET, T2I_ADAPTER, and IP_ADAPTER. The author also designed it to accommodate other custom types.

For example, you could select cpu for a connected module or cuda:0 for a CUDA device. This gives you control over device placement within a workflow. The device you enter still needs to be available in your ComfyUI environment; the override does not create a device or guarantee that every model will fit in its memory.

Caching CLIP Text Encode for FLUX

This node is intended for FLUX workflows that repeatedly encode prompt text. It remembers the CLIP model reference and the two text inputs, clip_l and t5xxx, so it can reuse a previous text-to-conditioning result when those inputs have not changed.

This is particularly useful when you keep a prompt and model the same while experimenting with other parts of a workflow. The cache responds to changes in either text input, a change of CLIP model, or cache expiry. It can retain entries for more than one CLIP model, so switching back to a previously used model does not necessarily mean losing its cached result.

The guidance value is not cached. You can change guidance without invalidating the text cache, which makes it easier to compare guidance settings while keeping the prompt unchanged.

A cache limit controls how many results are kept. When the limit is exceeded, the oldest cached result is removed. A smaller limit keeps fewer past results available; a larger one can be more useful when you revisit several prompts or models.

Caching CLIP Text Encode

Caching CLIP Text Encode offers the same basic time-saving idea in a simpler form. It has a single text input and no guidance value. Use it when your workflow calls for one text input rather than the FLUX node’s separate clip_l and t5xxx inputs.

As with the FLUX version, caching is most useful when you run a workflow repeatedly without changing the text or CLIP model. It avoids repeating text-encoding work; it does not change the words in your prompt.

Load Image From Path

Load Image From Path reads an image using its file path and returns three outputs: IMAGE, MASK, and the image path. The separate path output can help when another part of your workflow needs to know where the image came from.

The node also supports Krita .kra projects. It extracts the project’s mergedimage.png, allowing you to use the merged image without first exporting a separate PNG. If the loaded image has an alpha channel, the node detects it for mask generation. This can be useful when artwork contains transparent areas that you want to carry into an image-editing workflow.

Troubleshooting ComfyUI-ARZUMATA

A cached prompt does not seem to speed up a run. Check whether the text inputs and CLIP model are the same as on the earlier run. A changed input requires a new result. Also check whether the earlier entry has expired or was removed after the cache limit was reached. Caching only avoids repeated text encoding; other steps in the workflow still take time.

Changing guidance does not trigger new FLUX text encoding. This is expected. The FLUX node does not include guidance in its cache key, so you can adjust guidance while reusing the same encoded text.

A previous prompt or model is no longer in the cache. The cache removes its oldest entry when its configured limit is exceeded. Consider a higher limit if you regularly return to earlier prompts or CLIP models.

A device override does not work as expected. Confirm that the device name you entered refers to a device available to your ComfyUI setup. Then check that the connected model or module can run on that device. The override controls the requested placement; it cannot make unsupported hardware or insufficient memory available.

A Krita project is not loading the image you expected. The node uses the .kra file’s mergedimage.png. Check the merged appearance of the project in Krita if it differs from the individual layers you expected to use.

Learn More about ComfyUI-ARZUMATA

The original README identifies these projects as background for particular nodes:

These are related projects and credits, not required steps for using ComfyUI-ARZUMATA.

ComfyUI-ARZUMATA Related Nodes

RunComfy
Copyright 2025 RunComfy. All Rights Reserved.

RunComfy is the premier ComfyUI platform, offering ComfyUI online environment and services, along with ComfyUI workflows featuring stunning visuals. RunComfy also provides AI Models, enabling artists to harness the latest AI tools to create incredible art.

ComfyUI-ARZUMATA detailed guide | ComfyUI