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-OreX

ComfyUI Extension: comfyui-OreX

Repo Name

comfyui-OreX

Author
orex2121 (Account age: 2380 days)
Nodes
View all nodes(1)
Latest Updated
2026-11-14
Github Stars
0.07K

How to Install comfyui-OreX

Install this extension via the ComfyUI Manager by searching for comfyui-OreX
  • 1. Click the Manager button in the main menu
  • 2. Select Custom Nodes Manager button
  • 3. Enter comfyui-OreX 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-OreX Description

comfyui-OreX adds ComfyUI nodes for image editing, video and audio processing, prompt building, LoRA management, local AI assistants, canvas sizing, and workflow utilities in one visual workspace.

comfyui-OreX Introduction

comfyui-OreX is a growing collection of nodes for ComfyUI. It brings image editing, video and audio handling, prompt building, local AI assistants, and workflow utilities into the same visual workspace.

You can use it to crop and color-grade images, compare edits, make collages, prepare image sizes for generation, manage LoRAs, or divide a long sequence of frames into smaller pieces for processing. Some nodes connect to tools or models outside the extension, such as LM Studio, Ollama, and WAN SCAIL. Others work with images, text, or audio already moving through your ComfyUI workflow.

The author describes the collection as actively developing. This guide covers the nodes and behavior documented in the supplied information; it does not assume that undocumented nodes work in a particular way.

How comfyui-OreX Works

ComfyUI workflows connect nodes together: one node produces something, such as an image or a piece of text, and another node uses it. comfyui-OreX adds specialized steps you can place in those chains.

For example, you could use Ratio to choose a portrait canvas, generate an image, adjust it with Camera Raw, and use Image Compare to inspect the result alongside the original. Or you could use Image Chunk Cut to divide video frames into manageable groups, process each group, and reconnect them with Image Chunk Stich.

A few useful concepts make the nodes easier to approach:

  • A batch is a group of images or frames handled together. Chunk nodes divide a large batch into smaller groups when a later step cannot comfortably process everything at once.
  • A mask marks an area of an image for a workflow to use. Where supported, an OreX node can keep a mask aligned with an image as it is cropped or resized.
  • A latent is the working canvas used inside an image-generation model. Ratio can create empty latents in different formats; choose the output that matches the model family in your workflow.
  • A LoRA modifies how a compatible generation model behaves. The Lora Loader lets you arrange several LoRAs, adjust their strength, and decide which ones are active.

Some nodes also provide interactive controls directly on the ComfyUI canvas. You can drag a crop frame, compare images with a slider, or adjust a color edit in a larger visual editor instead of working only from numbers.

comfyui-OreX Features

Video and audio

Advanced Video Load accepts a video path or a file selected in the node. Its built-in player shows the source and, after the workflow runs, the processed result. You can choose among three audio behaviors:

  • Mute Original (Fast): remove the original audio without re-encoding the video.
  • Replace (Fast): use audio connected to new_audio while copying the video stream. Without a connected replacement, the source video and audio are copied unchanged.
  • Mix Audio (Re-encode audio): blend original and connected audio, with separate orig_vol and new_vol controls from 0.0 to 2.0. Without new audio, this mode can still adjust the original track’s volume.

Audio Load accepts an audio file or the sound track from a video. You can trim from the beginning and end using sliders or precise numeric fields. The built-in player stays within the selected range, so you can listen to the portion you will use. Its outputs include the trimmed audio, the file path used, and the trimmed duration. Selecting a different file resets the trim controls.

Scail is a specialized pose-guided video sampler for WAN SCAIL workflows. It processes a pose video in overlapping sections, carries context between sections, and joins the results into a longer sequence. Controls include pose strength and timing, reference-image inputs, section length and overlap, and standard sampling settings. It requires the WAN SCAIL nodes in ComfyUI; it is an extension of that workflow, not a standalone video model. The node reports its processing plan in the console and provides a progress bar.

Image editing and inspection

Camera Raw offers a full-screen color-editing view with a live preview. Its controls cover exposure, contrast, highlights and shadows, color temperature, saturation, texture, sharpening, blur, grain, vignette, and more. The HSL panel adjusts individual color ranges: for instance, you can change the saturation of blues without changing every other color. RGB Curves let you shape overall brightness and individual red, green, and blue channels. You can compare before and after, undo or redo edits, reset settings, and turn all adjustments off temporarily without losing their values. The node preview displays the original image; open the editor to see live adjustments.

Crop lets you drag a crop frame on the image preview or enter percentages to remove from each edge. Because those edge settings are percentages, the same crop can be used on images with different dimensions. You can lock an aspect ratio, choose a preset such as 1:1 or 16:9, enter a custom ratio, or set a target size in pixels or megapixels. Full Image, Center, and Maximize provide quick adjustments. If you connect a mask, Crop resizes it with the image. Its interpolation setting determines how the cropped image is resized; bicubic is the default.

Image Compare is a viewer for checking two images on the ComfyUI canvas. Use Slider for a draggable before-and-after boundary, Side-by-Side for a direct comparison, Overlap to blend the images, Difference to highlight pixel changes, or Blink to alternate between them. You can zoom and pan both views together. Its save control captures the current view as a JPG, or creates a looping GIF in Blink mode. This node is for viewing and saving comparisons; it does not pass the images onward to another node.

Sequences and collages

Image Chunk Cut divides a sequence of images, such as video frames, into equal-length groups. The default group length is 81 frames. If the final group is short, the node fills it with frames taken from a reversed portion of the sequence. It also reports how many frames it added as shortfall.

Image Chunk Stich reconnects groups in order. Connect the Cut node’s shortfall output to the Stich node’s trim_end input to remove the extra frames after processing. You can also trim frames from the beginning with trim_first. The result includes the joined sequence and its final frame count.

Image Merging creates a collage from 2 to 20 images. Choose a horizontal, vertical, or grid arrangement, including layouts that give one image more space than the others. You can choose which image guides the layout’s proportions, add a border, select a background-and-text color pair, and set an overall output size in megapixels. A multiline text field supplies captions in image order: one line for each image.

Image Merging V2 offers the same collage layouts and sizing controls, but gives each image its own captionN field. Choose V2 when captions come from different text nodes or when you would rather edit each caption beside its image input.

Prompts, local assistants, and LoRAs

Cip Text Encode builds a prompt from a main text field and additional connected strings. As you connect text inputs, new string1, string2, and later inputs appear automatically. Nonempty strings are placed before the main text, separated by commas. This makes it easier to combine, for example, a subject description and a selected style. The node outputs both the conditioning used for generation and the assembled prompt text, so you can inspect what was sent onward.

LM Studio connects a workflow to a locally running LM Studio server. You can enter a main prompt, add a system prompt, or select an editable system-prompt preset. Compatible vision models can also receive an image, which the node reduces to about 0.7 megapixels before sending. Controls include a seed, optional generation settings, whether to include the model’s reasoning text in the response, and options for freeing video memory before or after a request. The node outputs the response and a JSON record of the request it sent.

Ollama serves a similar purpose for a locally running Ollama server. It shares the system-preset file used by the LM Studio node, while using Ollama’s own model list, generation options, and keep_alive behavior for how long a model remains in memory. Both nodes depend on a suitable local server and model; the extension does not supply the language model.

Lora Loader manages a stack of LoRAs. Add rows, choose files, change each LoRA’s strength in 0.05 steps or by entering a value, reorder rows, and switch individual LoRAs or the whole stack on and off. You can edit trigger words and decide whether each row contributes them to the combined trigger-word output. The node looks for local metadata and previews first; if it cannot find them locally, it can look up a LoRA by file hash through the Civitai API and cache the result. Missing LoRA files are highlighted and stop execution with an explanatory error, while disabled rows are excluded from that check. Outputs include the modified model and CLIP, active trigger words, and names of the used LoRA files.

Canvas sizing

Ratio calculates image width and height from an aspect ratio and target megapixel setting. Choose from presets or adjust a visual rectangle for a custom shape. Its Multiplicity setting rounds dimensions to a chosen multiple, which can help match a model’s size requirements. Ratio also produces three different empty latent outputs; use the one appropriate for your generation model.

comfyui-OreX Models

comfyui-OreX is primarily a node collection, not a single AI model. The documented nodes work with several model families or external model services:

Model or serviceWhere it appearsWhen to use it
SD1.5 / SDXLRatio’s standard empty latent outputWhen preparing a canvas for a compatible SD1.5 or SDXL workflow.
SD3 / Flux 1Ratio’s SD3/Flux 1 empty latent outputWhen your workflow uses one of these model families.
Flux 2Ratio’s Flux 2 empty latent outputWhen preparing dimensions and a latent for a Flux 2 workflow.
WAN SCAILScailWhen generating pose-guided video using the required WAN SCAIL nodes.
LM Studio modelsLM StudioWhen you want a locally served language model to produce workflow text, or a compatible vision model to respond to an image. The available models come from your LM Studio server.
Ollama modelsOllamaWhen you want to use a model available on your local Ollama server.
damo/cv_unet_skin_retouching_torchSkin Retouching AIFor AI-assisted skin retouching. The documented strength control blends the original image with the model’s result.

The correct latent output in Ratio depends on the model family you are using. Changing a collage layout, crop method, or Camera Raw setting changes how an image is prepared or displayed; those settings do not switch the underlying generation model.

Troubleshooting comfyui-OreX

A saved workflow says a LoRA is missing.

  1. Open Lora Loader and find the row marked with a warning.
  2. Select an available LoRA file for that row, or disable the row if you do not need it.
  3. Run the workflow again. Active rows pointing to missing files are blocked; disabled rows are not included in that check.

My reassembled sequence contains extra frames at the end.

  1. Check the shortfall value from Image Chunk Cut.
  2. Connect it to Image Chunk Stich as trim_end.
  3. Run the workflow again to remove the frames added to complete the final group.

Image Chunk Stich returns only one frame after trimming.

Check trim_first and trim_end. When their combined values would remove the entire sequence, the node warns in the console and returns one frame rather than failing. Reduce the trim values and run it again.

Advanced Video Load does not replace the sound.

Confirm that an audio source is connected to new_audio and that Replace (Fast) is selected. In Replace mode, leaving new_audio disconnected copies the existing video and sound unchanged.

Audio Load is playing or producing the wrong section.

Check both the sliders and their corresponding precise trim fields. A nonzero precise field takes priority. You can change the values directly, or select the file again to reset all four trim controls to zero.

A Scail workflow cannot run.

Confirm that the WAN SCAIL nodes are available in your ComfyUI setup. Scail relies on comfy_extras.nodes_scail; it does not replace those underlying nodes.

Frequently asked questions

Can I use Image Compare to feed an image into the next node?
No. Image Compare displays images and can save the current comparison view, but it does not forward image data.

What is the difference between Image Merging and Image Merging V2?
Both create the same kinds of collage. The original uses one multiline caption field; V2 gives each image a separate caption field that can also receive a string connection.

Why does the Camera Raw node preview show my original image?
Its node preview intentionally keeps the original. The full-screen editor applies live adjustments over that source, so you can inspect your changes there.

Do LM Studio and Ollama provide models through comfyui-OreX?
No. These nodes connect to the respective local servers and use models available there. Their shared presets help organize instructions, but do not install or replace a model.

Learn More about comfyui-OreX

The nodes include controls designed to help you learn while working. Camera Raw, Crop, LM Studio, Ollama, and Ratio provide Russian and English parameter hints; Camera Raw also has guidance inside its full-screen editor. For visual tasks, the built-in viewers in Advanced Video Load, Audio Load, Crop, and Image Compare let you check results without leaving the workflow.

For LoRA examples and metadata, Lora Loader can display locally available previews and, when needed, retrieve information through the Civitai API. Check that a LoRA and its trigger words suit the model and artwork you are using before adding them to a finished workflow.

comfyui-OreX 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-OreX detailed guide | ComfyUI