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-SaturnNodes

ComfyUI Extension: ComfyUI-SaturnNodes

Repo Name

ComfyUI-SaturnNodes

Author
kofiblto (Account age: 1281 days)
Nodes
View all nodes(1)
Latest Updated
2026-11-08
Github Stars
0.0K

How to Install ComfyUI-SaturnNodes

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

ComfyUI-SaturnNodes adds visual LoRA and image browsing, prompt batching, live previews, queue pause and recovery, image utilities, and stereoscopic 3D output to ComfyUI workflows.

ComfyUI-SaturnNodes Introduction

ComfyUI-SaturnNodes is a collection of nodes and interface tools for managing creative workflows in ComfyUI. It brings together visual file browsing, prompt batching, live generation previews, queue controls, image utilities, and stereoscopic 3D output.

If you make many variations of an image, work with a large LoRA collection, or leave generations running while you do something else, SaturnNodes can reduce repetitive work. For example, you can prepare several prompts in one text box, send them through a queue one at a time, watch a generation develop, and recover unfinished queue items after a restart.

Most features support both ComfyUI Frontend V1 (LiteGraph) and Frontend V2 (Vue UI). The Visual LoRA Loader currently breaks on Frontend V1, so use Frontend V2 for visual LoRA browsing. For guided examples, see the Complete Walkthrough.

How ComfyUI-SaturnNodes Works

ComfyUI workflows are graphs: each node receives information, performs a task, and passes its results to another node. SaturnNodes adds specialized steps to that graph and, for some features, controls in the ComfyUI interface.

A few examples show how the pieces fit together:

  • A prompt queue works like a stack of creative briefs. Write several prompts, choose how they are separated, and let Prompt Queue Iterator select one for each queued run. Prompt Counter shows how many prompts you have prepared.
  • A LoRA loader works like choosing an accessory for your base model. Select a LoRA by name, thumbnail, or a tag in your prompt. The loader applies it to the model and CLIP connections that continue through your workflow.
  • A depth map works like a rough guide to distance. The Stereoscopic 3D / Cross-Eye Generator uses that guide to shift parts of an image into slightly different left-eye and right-eye views. You can provide a depth map or let the node create one automatically.
  • Queue controls act like a checkpoint. You can pause execution, place a decision step inside a workflow, or restore unfinished items after an interruption.

Some settings belong to individual nodes, while others are available from the ComfyUI Settings gear icon. Visual loaders, preview docks, queue behavior, and optional online preview searches can be adjusted there.

ComfyUI-SaturnNodes Features

LoRA Selection and Browsing

LoRA Loader (Folder) lets you select a LoRA from a chosen subfolder, such as Characters or Styles/Anime. Adjust strength_model and strength_clip to control how strongly it affects each connection. Its name output can be formatted as a readable name, filename, path, or custom regex result, which is useful when naming or recording generated images.

LoRA Loader (Pretty) offers the same basic loading task with more readable names in its dropdown. A file named char_Cyberpunk-Character_v1.safetensors, for example, can appear as Cyberpunk Character V1.

Visual LoRA Loader presents LoRAs as thumbnails and supports multiple selections. You can resize its tiles for that particular node, switch between a scrollable grid and showing all items, and sort by name, modification date, or usage. Optional Civitai and TMDB preview searches are disabled by default and can be enabled in Settings. Usage tracking adds ranking badges and can be turned off.

The Visual LoRA Loader also has three confirmation-based maintenance controls:

ControlWhat it doesWhen to use it
↺ CounterResets the node’s sequential or random selection position and its prompt-to-LoRA assignments.Between batches only. Resetting while workflows are running or queued can pair prompts with different LoRAs than intended.
🧹 UsageClears LoRA usage counts and badges.When you want to start usage-based ranking over; it does not change LoRA loading.
🗑 CacheDeletes cached thumbnails, which are recreated when you view the grid.When previews appear stale or you want them regenerated.

Text LoRA Finder & Loader reads prompt text for tags such as <lora:name:strength> or names matching LoRAs in a selected folder. It applies matching LoRAs and can remove the matched tags from the text it passes onward. Fallback model and CLIP strengths are available when a tag does not specify a strength.

Images and Live Previews

Visual Image Loader displays images from a folder as selectable thumbnails. It outputs the selected image, its dimensions, and any positive-prompt metadata it can extract. You can change tile size, display mode, and sorting. For safety, its folder access is limited to ComfyUI’s input, output, and temp directories. By default, it checks the newest 128 images for prompt metadata; the limit can be changed in Settings, or set to 0 to disable that scan.

Load Image From Folder watches a folder in ComfyUI’s input or output directory for an image to load. You can tell it to wait when the folder is empty or continue with a dummy image, set how often it checks, sort candidate files, and filter filenames with a regular expression. Its has_image output tells downstream nodes whether a real image was found. Leave delete_image off unless you intend to remove the source file after it loads.

Recent Outputs loads from a pool of the newest images in an output folder. Set amount from 1 to 100, then use index to step through that pool.

Live Latent Preview displays an in-progress preview inside the graph while a sampler generates an image. You can also enable a floating preview dock in Settings. The dock can hide when idle, and its maximum width and height are adjustable by dragging its top-left corner or changing the settings.

Copy Prompt actions can be shown on generated images and in node right-click menus. Settings also include an optional Inspect Asset button for image cards in the Assets pane. These are interface conveniences; they do not change the generated image.

Prompts and Batch Iteration

Prompt Counter accepts multiline text and shows the number of prompts as you type. You can separate prompts by newlines, groups of empty lines, or a custom regular expression. Its text output remains unchanged, so you can connect it to Prompt Queue Iterator or another text input.

Prompt Queue Iterator selects a prompt block for each queued run. Choose:

  • Sequential (Loop on End) to return to the start after the last prompt.
  • Sequential (Stop on End) to remain on the last prompt once the queue is exhausted.
  • Random (Keep) to make random selections without consuming an ordered queue.
  • Random (Cycle) to work through a randomized cycle.

The iterator reports the selected prompt, remaining text, remaining count, and whether a stop-on-end queue is exhausted. By default, Text & Node tracking keeps a separate position for each prompt text and separator configuration. Choose Node Only if you want the node’s position to persist through prompt edits or separator changes. Reset Counter (0) also clears that node’s saved recovery assignments and random shuffle order.

Multi Text Replacer applies several text substitutions in one step. Supply matching find and replacement lists, separated by commas or lines. You can enable case-sensitive matching or regular expressions when simple exact phrases are not enough.

Text Split divides text into two outputs at a chosen delimiter. It can split at the first occurrence from the start or the last occurrence from the end, and it can trim whitespace around the results. For example, splitting portrait -- cinematic lighting at -- gives you two text parts to use elsewhere in a workflow.

For queued batches, Batch Graph Snapshot Guard is enabled by default. It captures prompt text, LoRA selections, and node inputs when you click Queue so later edits to the canvas do not alter those queued items; random seeds can still randomize. Optional colored lines in the queue list help you see which items belong to the same batch.

Queue, Pause, and Recovery Controls

Saturn Decision pauses a workflow at a chosen point and presents Continue, Cancel, or Stop Workflow. Its boolean cancel output can guide a downstream branch. You can bypass the decision with disable, request an operating-system notification, or set a timeout in seconds; -1 means it will wait indefinitely.

SaturnNodes also adds queue-level pause and continue controls. In Settings, you can choose whether a pause finishes the active prompt or takes effect at the next resume point, whether the queue starts paused, and whether your pause state survives a restart. An optional system-tray icon provides queue status and controls outside the browser.

Persistent Queue is enabled by default. It saves unfinished queue items and attempts to restore them after a restart or crash. You can choose whether recovered items follow your default launch state, start paused, or start running, and set how many additional launches may retry an item that fails to restore.

Assets & History Restore can repopulate the Assets pane with recent output images on launch. By default, it includes the newest 64 images and searches output subfolders. Restore Now starts a rescan without restarting ComfyUI. Generation time is shown where SaturnNodes has a recorded duration; older images without one do not receive an invented duration.

Image Size and Aspect Ratio

Text Aspect Ratio Finder looks for a ratio such as 16:9 or 2.35:1 in your text, then calculates width and height for a target megapixel count. You can limit it to a list of allowed ratios, choose the first or last match, set a fallback ratio, and constrain the result with minimum and maximum megapixels or a dimension multiple. It also outputs the text with the ratio token removed. Leave the allowed-ratios field completely empty to accept any valid ratio it finds.

Preview Image Size & Aspect Ratio reports the width, height, and closest matched aspect ratio of an input image. Use it when you want to check image dimensions without calculating them yourself.

Stereoscopic 3D and Cross-Eye Images

Stereoscopic 3D / Cross-Eye Generator turns one 2D image into left-eye and right-eye perspectives using a depth map. Its output formats include Cross-Eye, Parallel (Wall-Eyed), Red-Cyan Anaglyph, Side-by-Side (Parallel), Top-Bottom (Over-Under), and individual eye views. It also outputs the eye images and the depth map separately, so you can inspect or reuse them.

Start with the default settings, then adjust the image:

  • depth_intensity controls the strength of the separation between perspectives. A higher value makes the 3D effect stronger.
  • convergence_plane chooses the depth that appears to sit at the screen surface. Objects on either side of it can appear to project outward or recede.
  • depth_gamma changes how depth values are distributed; invert_depth swaps near and far.
  • Zoom and pan let you frame the part of the artwork you want before the 3D views are made.
  • Gap spacing, outer border, and alignment dots can make paired images easier to view, particularly for cross-eye viewing.
  • fill_method selects either smooth bilinear filling or nearest-neighbor filling for gaps created when pixels shift.
  • preview_zoom changes the canvas preview size without requiring you to change the intended viewing format.

You may connect your own depth_map. If you leave it unconnected, the node uses an automatic Depth Anything V2 Small ONNX fallback, downloaded on demand.

Local Scripts, Safety, and Personalization

Run Local File can run a script or executable as part of a workflow. This is an advanced feature for tasks such as connecting a generation to an existing local processing script. Execution is off by default in Settings and is restricted to files inside ComfyUI/scripts/; absolute paths and .. path traversal are blocked. Each run also needs the operator to click ⚡ Authorize Run (5 min) on the node. That authorization is single-use and is not stored in shared workflows or image metadata. Only authorize scripts you understand and trust.

Other Settings let you choose custom SaturnNodes colors, hide individual nodes from the Add Node menu, and export a debug profile for troubleshooting. Hiding a node requires a server restart to take effect but does not unregister it from existing workflows. Optional Civitai and TMDB searches are off by default. Local settings and API keys are stored in the SaturnNodes user area; the debug profile excludes sensitive keys and tokens.

ComfyUI-SaturnNodes Models

SaturnNodes is primarily a workflow and utility extension, not a collection of interchangeable image-generation checkpoints.

  • Your own LoRAs are used by the folder, pretty-name, visual, and text-based LoRA loaders. Which ones are available depends on the LoRA files in your ComfyUI collection. A character LoRA and a style LoRA, for example, will influence different aspects of a result; their model and CLIP strengths let you adjust how strongly they are applied.
  • Depth Anything V2 Small ONNX is the named model used by the stereoscopic node when you have not connected a depth map. It estimates the image’s depth so the node can create two perspectives. If you already have a depth map you prefer, connect it instead.

The provided documentation does not identify additional bundled models or a model-selection menu for the stereoscopic fallback.

Troubleshooting ComfyUI-SaturnNodes

The Visual LoRA Loader does not work in Frontend V1

  1. Check whether you are using Frontend V1 (LiteGraph).
  2. Open the workflow in Frontend V2 (Vue UI) for visual LoRA browsing.
  3. If you need to remain on V1, use a nonvisual LoRA loader for now. The documented V1 issue specifically affects VisualLoraLoader.

A visual image folder shows no images

  1. Confirm that the folder path points inside ComfyUI’s input, output, or temp directory.
  2. Check that the folder contains the images you expect.
  3. Review the node’s sort setting if the image you want is not where you expected it in the grid.

For Load Image From Folder, paths are restricted further to input and output. If it is waiting indefinitely, check the watched folder and the wait_if_folder_is_empty setting. If has_image is false, the node did not load an image from that check.

A prompt queue produces unexpected prompts

  1. Compare the text in Prompt Counter or Prompt Queue Iterator with the selected separator. A newline separator treats text differently from a separator based on multiple empty lines.
  2. If using Custom Regex, check that the pattern describes the actual delimiter in your text.
  3. Check pop_mode: looping, stopping, and the two random modes intentionally behave differently.
  4. Check track_by if you recently edited prompts. Text & Node and Node Only retain position differently.
  5. When no batch is running or queued, use Reset Counter (0) to begin a fresh cycle. For all queues, Settings offers Reset All Queues.

A queued prompt uses an unexpected LoRA

Check the Visual LoRA Loader’s selection mode and whether ↺ Counter was reset during an active or queued batch. That reset clears prompt-to-LoRA assignments and can restart the selection cycle. Finish or clear the affected batch before resetting again.

Thumbnails or online previews are missing

  1. Check whether the relevant Civitai or TMDB Auto-Scraping setting is enabled; both are off by default.
  2. If the online service needs a credential for your use, check the corresponding key or token in Settings.
  3. Use 🗑 Cache on the Visual LoRA Loader to regenerate its cached thumbnails when you next open the grid.
  4. If a previous online lookup failed, use Clear Scrapes Cache in Settings to allow another attempt.

The Disable Thumbnail Downscaling (Serve Raw) setting serves original preview images instead of faster-loading downscaled WebP thumbnails. Leave it off unless you specifically need raw previews.

The local file node will not run a script

  1. Confirm Allow Local File Execution is enabled in Settings.
  2. Place the script inside ComfyUI/scripts/ and enter a path relative to that directory. Absolute paths and paths containing .. are not permitted.
  3. Click ⚡ Authorize Run (5 min) on the node before the run; authorization expires and is used only once.
  4. Inspect stderr, exit_code, and success for the result. A blocked or timed-out run can report -1.

Frequently Asked Questions

Do I need Civitai or TMDB accounts to use SaturnNodes?
No. Their automatic online preview searches are optional and disabled by default. The local SHA256 hash search described for LoRA previews remains available independently of Civitai auto-scraping.

Will hiding a node remove it from an existing workflow?
No. Node-management toggles hide nodes from the Add Node menu after a server restart; disabled nodes remain registered so existing workflows can still load.

Does resetting LoRA usage change my images?
No. 🧹 Usage clears counts, badges, and usage-based ordering, not the way LoRAs load. The separate ↺ Counter control does affect selection order and should be used between batches.

Can I use my own depth map for 3D output?
Yes. Connect it to the stereoscopic node’s optional depth_map input. The automatic Depth Anything V2 Small fallback is used when that input is not connected.

Can SaturnNodes recover work after ComfyUI closes unexpectedly?
Its Persistent Queue feature is enabled by default and attempts to restore unfinished queue items after a restart or crash. Settings control the restored queue’s launch state and retry behavior.

Learn More about ComfyUI-SaturnNodes

  • Complete Walkthrough: Step-by-step node recipes and batch-prompting patterns. Start here when building your first SaturnNodes workflow.
  • Custom Node Security & Compliance Rulebook: More detail on the extension’s security design, especially relevant if you plan to use script execution or administrative controls.
  • MIT License: The terms under which the extension is distributed.

ComfyUI-SaturnNodes 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-SaturnNodes detailed guide | ComfyUI