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 > Fantastic Loras

ComfyUI Extension: Fantastic Loras

Repo Name

comfyui_fantastic-loras

Author
adudeguyman (Account age: 1113 days)
Nodes
View all nodes(1)
Latest Updated
2026-10-17
Github Stars
0.02K

How to Install Fantastic Loras

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

Fantastic Loras Description

Fantastic Loras helps ComfyUI users filter, stack, and save LoRAs across model paths, then compare their effects at different strengths through automated image grids and an interactive viewer.

comfyui_fantastic-loras Introduction

comfyui_fantastic-loras is a LoRA toolkit for ComfyUI. A LoRA is a small add-on that changes how a compatible image-generation model behaves—for example, by introducing a character, style, or detail. If you work with many LoRAs, this extension helps you find the ones you want, organize them, and compare their effects without building a long chain of loader nodes.

The toolkit can help you:

  • Keep your choices manageable: show LoRAs from selected folders instead of searching your entire collection.
  • Build a LoRA stack in one place: add up to 12 LoRAs to one loader, control their order and strengths, and apply them to as many as five model paths.
  • Save combinations: store a stack as a named preset and bring it into another workflow.
  • Compare results fairly: test different LoRAs and strengths in an XY comparison with one Queue press.
  • Reuse selections across model paths: experimentally mirror LoRAs configured in another loader onto a separate model.

The extension includes a grid saver and an interactive viewer, so you can either keep a labelled comparison image or inspect individual results on the ComfyUI canvas. Its panels work with both the classic canvas and Nodes 2.0.

Fantastic Lora Loader showing its slots and presets

How comfyui_fantastic-loras Works

Think of your base model as a starting canvas and each LoRA as an adjustment you can add to it. A strength controls how strongly that adjustment is applied. You can use several LoRAs together, but the order and strength of each one can affect the result.

The extension offers two main ways to work:

  1. Stack LoRAs with the Fantastic Lora Loader. Each filled slot adds a LoRA to the model. The slots are applied in order, so you can drag them into the sequence you want. The loader can feed several model paths from the same panel. For example, a detail LoRA could be set to 1.0 on a base pass and 0.4 on a refiner pass.
  2. Test LoRAs separately with the Fantastic Lora Plotter. Here, slots are candidates for comparison, not layers to stack. If you add three LoRAs and test each at two strengths, the Plotter prepares six test images, plus any baseline images you enable. It sends the tests through the downstream workflow from a single Queue press.

For a meaningful comparison, keep the seed fixed so the starting noise does not change between cells. Otherwise, you may be comparing changes caused by the seed as well as changes caused by the LoRA. Check the Plotter’s SWEEP footer before queueing: the number of generations grows as you add LoRAs, strengths, and model paths.

Most nodes also provide a FOLDERS filter. It is like narrowing a file browser to the shelves you are currently using: the LoRA picker and randomizer then draw from those selected folders.

comfyui_fantastic-loras Features

Fantastic Lora Loader 📁

The main loader holds up to 12 LoRA slots. Connect a MODEL input and pass its output to your sampler. Connect CLIP as well if your setup uses LoRA weights for its text encoder; for model-only LoRAs, CLIP can remain unconnected. Click Add lora… to search for a LoRA, or click an empty slot.

Each slot lets you:

  • Enable or disable a LoRA without removing it.
  • Adjust its strength using the stepper controls, mouse wheel, or a typed value. A disabled or zero-strength LoRA is skipped.
  • Swap or remove the selected LoRA.
  • Drag it into a different position to change application order.
  • Open ⚙ to choose which model paths receive it and set a different strength for each path.

The chains control adds model inputs and outputs, up to five. New LoRAs are routed to every model by default. If you change an individual model’s strength in ⚙, it stops following the slot’s shared strength. The footer shows which LoRAs each model output will receive, in order, so you can check the result before running a workflow.

Use 🎲 Add random to make a slot pick a LoRA from the allowed folders. You can roll immediately, lock the current pick, or enable auto-roll for a fresh selection each time you queue. A random slot can have a narrower folder scope than its node, but it cannot choose from folders excluded by the node’s filter.

Per-model LoRA routing and strength controls

Folder filtering, favourites, and presets

The FOLDERS bar limits what appears in the picker and what random slots can choose. Search for a folder, star frequently used folders, or select a branch of the folder tree. Choosing one folder by itself includes LoRAs directly inside it; use the branch toggle if you also want its subfolders. The picker re-reads your folders when you open it.

You can star individual LoRAs to bring them to the top of the LoRA picker. The loader also offers four panel themes and an .ext control to show or hide .safetensors in displayed names.

Presets save a named loader setup so you can return to it later:

  • Save stores the current stack and asks before overwriting an existing name. For random slots, you can choose whether to preserve their random behavior or freeze their current picks.
  • Load replaces the current stack, folder filter, and model count.
  • + Add to stack brings a preset’s LoRAs into the current stack without replacing it. Duplicates are skipped, and the loader reports anything that cannot fit in the 12 slots.
  • Categories and favourites help organize presets by project, model, or another grouping that makes sense to you.

A loaded preset is marked “(modified)” if you change its saved setup. Presets are shared across workflows, while a workflow also saves the LoRA configuration currently on its nodes.

Folder picker with searchable subfolders

Fantastic Lora Plotter 📊

The Plotter is an XY test bench for seeing what a LoRA actually contributes.

  • Per-line strength: test each selected LoRA once at the strength shown on its slot. Use this to shortlist favourites.
  • Global strengths: test every selected LoRA at every strength in a shared list, such as 0.5, 0.75, 1.0. Use this to compare both LoRA choice and intensity.
  • Control image: add an image with no LoRA as a reference to the base model.
  • Multiple model chains: run the same comparison across up to five model paths.

Connect the Plotter’s MODEL output to the rest of your generation workflow and send its metadata output, together with the generated images, to the Fantastic Plotter Image Saver. One Queue press starts the full comparison, but downstream steps run for every test cell. If your workflow includes an expensive upscaler after the Plotter, that upscaler also runs for every cell.

For example, four LoRAs × three strengths = 12 test images on one model path, before control images. Across two model paths, that becomes 24 test images before controls.

Plotter configured for a LoRA strength comparison

Fantastic Plotter Global Lora 🌐

Use this node for LoRAs that should stay constant across every Plotter test. If you are comparing character LoRAs but want the same painterly style in every image, put the style LoRA here. Each enabled global LoRA has its own fixed strength.

It also offers two separate baseline options: a pure base-model image and an image with only the global LoRAs applied. These make it easier to distinguish the effect of the LoRA under test from the effect of your constant stack. When this node is connected, its baseline controls take over from the Plotter’s own control-image toggle.

Fantastic Plotter Image Saver 📊

The Image Saver combines the generated cells into one labelled grid image that you can pass to a Save Image node. It also passes the individual images through for the Grid Viewer.

Choose between:

  • Overlay: a label appears on each cell, so individual images remain identifiable when cropped or shared.
  • Classic: LoRA names appear along the left and strengths along the top, leaving the cells unmarked. This needs a complete LoRA-by-strength comparison; if the results do not form one, the saver falls back to Overlay.

You can adjust label text, background, font size, padding, and—when using Overlay—label opacity. Constrain Image Output Size can shrink each cell before assembly to keep a large grid more manageable. Automatic column detection normally arranges a Global-strength sweep with LoRAs as rows and strengths as columns.

Overlay and Classic comparison-grid layouts

Fantastic Plotter Grid Viewer 🔍

The Grid Viewer lets you explore a comparison on the canvas. Connect the Image Saver’s individual images output and matching metadata, or use the saver’s 🔍 Add Grid Viewer (connected) button. Do not connect the composed grid image: it no longer contains separate cells for the viewer to rearrange.

You can zoom into a cell, hide and restore rows or columns, adjust thumbnail size, and select images for a side-by-side comparison. Comparisons can be exported as a PNG or saved for reopening with that run’s grid.

The latest grid can reappear when you reopen a workflow during the same ComfyUI session. To keep it after a restart, use Save Grid or enable auto-save every run. Saved grids can be browsed and loaded; pinned grids are protected from automatic cleanup. Auto-save uses disk space, so review the archive settings if you generate many comparisons. A new Plotter run replaces the viewer’s previous grid and its selections and saved comparisons.

Interactive Grid Viewer showing comparison images

Fantastic Lora Mimic 🪞

The Mimic is experimental. It applies LoRAs configured in another loader to the Mimic’s own MODEL and CLIP path. This is useful when two passes need related LoRA selections but should not share an already-modified model.

When you pick a compatible source node, the Mimic can follow its selections live. Individual mirrored LoRAs can either stay linked to the source strength or be unlinked for an independent strength. Supported source families include the extension’s stack nodes, stock ComfyUI LoRA loaders, rgthree’s Power Lora Loader, and specified Efficiency and Comfyroll stackers. You can also connect a compatible LORA_STACK output; this wire-based route mirrors the resolved stack but does not provide the same per-LoRA strength controls.

High / Low Model Mode is intended for paired LoRAs used with split model paths, such as Wan 2.2 high- and low-noise passes. It helps find a matching companion LoRA for the Mimic’s path and lets you set that companion’s strength. A Subgraph Companion can forward compatible LoRA selections across a subgraph boundary, although forwarded stacks do not provide the Mimic’s companion-selection controls.

Because the Mimic reads other nodes’ settings through unofficial graph interfaces, check its displayed selections before starting a long generation run.

Mimic mirroring LoRAs from another loader

Fantastic Any Selector 🎯

The Any Selector is a filtered filename picker, not a file loader. It can supply a selected filename to a compatible loader input for categories such as models, CLIP, VAEs, or upscale models. Right-click a loader and choose 🎯 Add Fantastic Any Selector to place and connect one.

Its folder filters, favourites, and presets are kept separately for each detected file category. It supports one target connection at a time so it knows which category of files to display. If it is unwired, it cannot determine a category. Because its output can connect to many input types, make sure you connect it to an appropriate filename input; an unsuitable connection may fail when queued.

Fantastic Seeds 🌱

Fantastic Seeds supplies seeds using fixed, randomize-every-queue, or locked-random modes. It also keeps a history of the last 10 queued seeds, which can help you revisit a promising result. For LoRA comparisons, use a consistent seed across test cells.

What's New with comfyui_fantastic-loras

The original documentation identifies the current release as 2.2.0 and highlights these changes:

  • Version 2.2: Added Fantastic Seeds 🌱, with three seed modes and a 10-seed history. The folder picker became a collapsible tree, making nested collections easier to navigate. The Loader and Plotter footers also show how many LoRAs pass the current folder filter.
  • Version 2.1: Added Fantastic Any Selector 🎯 for filtered filename selection across loader categories, including per-category presets. The Grid Viewer gained clearer placement for hidden-row and hidden-column labels, and overlapping extension panels now stack correctly when clicked.

Moving from version 1: The older single-model Fantastic Lora Loader was replaced by the multi-model loader. Existing workflows using the old node need that node swapped for the new one; their LoRA list, strengths, and folder filter carry over. Version 1 favourites and theme choices were stored in the browser and do not migrate to the newer settings, so you will need to star favourites again.

Troubleshooting comfyui_fantastic-loras

A node panel is blank, buttons do nothing, or the display looks wrong after changing canvas renderers.

  1. Hard-refresh your browser with Ctrl+Shift+R to clear cached interface code.
  2. If the issue followed an update, restart ComfyUI and refresh again.

Presets or favourites are not saving.

  1. Restart ComfyUI, rather than only refreshing the browser. The extension adds server routes when ComfyUI starts.
  2. Reopen the workflow and try saving again.

A new LoRA folder does not appear in your choices.

  1. Open the node’s FOLDERS picker; it re-reads the folders when opened.
  2. If you already selected specific folders, enable the new folder or its branch.
  3. Check that the node’s filter includes the folder before searching for its LoRAs.

A preset or workflow refers to a LoRA that is no longer on disk.

  1. Check the preset-load notice or ComfyUI console for the missing name.
  2. Choose an available replacement, or restore the missing file if you still want it. Missing LoRAs are skipped rather than stopping the entire workflow.

The Plotter is generating more images—or a larger grid—than expected.

  1. Read the SWEEP footer before queueing.
  2. Count LoRAs × strengths × model chains, then account for enabled controls.
  3. Reduce the number of tests, or turn on Constrain Image Output Size in the Image Saver if the composed grid image is too large.

The Grid Viewer does not show individual comparison cells.

  1. Connect the Image Saver’s images output, not its flattened grid output.
  2. Connect the matching metadata, or use Add Grid Viewer (connected) to create the connections automatically.

A grid disappears after a full ComfyUI restart.

By default, the viewer’s grid images are in ComfyUI’s temporary folder. Use Save Grid or enable auto-save every run to keep future grids on disk. Saved grids still require their associated output files to remain available.

The Classic grid layout switches to Overlay.

Classic needs a complete LoRA × strength rectangle. Use the Plotter’s Global strength sweep to produce that layout; otherwise, use Overlay labels.

Frequently asked questions

Do I need to connect CLIP?
Connect it if your workflow uses text-encoder LoRA weights. For model-only LoRAs, the model path can be used without CLIP.

Does adding several LoRAs to the Plotter stack them together?
No. Plotter slots are separate tests. Use the Loader to stack LoRAs, or Global Lora to keep selected LoRAs active across every Plotter test.

Can I use different LoRA strengths on different model paths?
Yes, in the Fantastic Lora Loader. Open a slot’s ⚙ options to adjust its routing and per-model strengths. The Plotter instead tests each selected LoRA across its connected model paths.

Can I change the Mimic’s mirrored LoRA strengths?
When you choose a source in its interface, you can unlink an individual LoRA and set its strength independently. With a wired LORA_STACK, change the strength upstream instead.

Learn More about comfyui_fantastic-loras

  • Example workflows: The extension includes two ready-made workflows in its example_workflows folder. Drag one onto the ComfyUI canvas to explore its connections and settings.
  • Grid Viewer interaction demo: Open grid_viewer_demo.html in a browser to preview the viewer’s interactions outside ComfyUI.
  • Extension repository: Browse the comfyui_fantastic-loras project for the included examples, screenshots, and documentation.

Fantastic Loras 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.

Fantastic Loras detailed guide | ComfyUI