ComfyUI_LC123_nodes Introduction
ComfyUI_LC123_nodes is a collection of custom nodes and canvas tools for ComfyUI. It helps you organize complicated workflows, choose model files, apply LoRAs, and finish images or videos without building every adjustment from scratch. The author’s supplied README describes version 1.56.10 as containing 160 Python nodes and five JavaScript-only tools.
The collection covers several kinds of work:
- Workflow organization: alignment guides, labels, notes, quieter previews, and tools for finding performance bottlenecks.
- Model preparation: image and video model loaders that recommend files and compatible speed-ups for your machine.
- Creative control: LoRA loaders, image relighting, skin work, sharpening, color matching, film effects, and depth-based effects.
- Review and sharing: before-and-after previews, translated workflow notes, and reports you can inspect without changing your generations.
You do not need every feature in one workflow. For example, you might use Align to tidy a graph, LC Model Optimizer to prepare its model, and LC Natural Look to soften an overly glossy final image.
For the complete node and category list, see NODES.md. The sections below focus on the features described in the supplied documentation.
How ComfyUI_LC123_nodes Works
A ComfyUI workflow is a set of connected steps. One node loads a model, another generates an image, and later nodes adjust or save the result. LC123 adds tools at different points in that process:
- Before generation, the optimization report checks what your machine can run. LC Model Optimizer can then select suitable model files and supported speed-ups. Think of the report as a guide and the optimizer as a loader that puts its recommendations to work.
- During workflow building, Align guides help you place nodes neatly. LoRA loaders let you arrange several style or character additions in one place instead of scattering loaders across the graph.
- After generation, image nodes can change the look of an existing result. LC Lighting Control V2, for instance, uses maps describing surface direction and distance to calculate new light and shadows. It adjusts the supplied image rather than generating a new scene.
- While reviewing, previews let you compare an effect with the original. On supported image-effect nodes, click the preview to zoom to a detail and move the pointer to wipe between before and after.
Canvas appearance is separate from image quality. Settings such as half-resolution previews reduce what ComfyUI draws on the canvas; they do not reduce the image passed between nodes. By contrast, settings on image-processing nodes can change the final output.
ComfyUI_LC123_nodes Features
Canvas organization and reports
Align adds rulers and movable guide lines. Left-click a ruler for a node guide or right-click for a group guide; drag a marker to move its line and click it again to remove it. Nodes snap to node guides, groups snap to group guides, and LC Notes can snap to either. Guides and the Align on/off state are saved with the workflow. The snap distance follows ComfyUI’s Snap to grid size, so a smaller grid gives you finer placement. You can disable the tool under Settings > LC Settings ⚙️ > Align.
Align button, rulers, and canvas guides
Connection FX highlights valid sockets while you drag a wire, making crowded graphs easier to connect. It is off by default and works on the classic canvas.
Three read-only reports answer different questions:
| Report | Use it when you want to know… | What to expect |
|---|---|---|
| Comfy Optimization Report | Why the open workflow’s canvas feels slow | A prioritized list of relevant pack, LC123, ComfyUI, and sidebar settings |
| System & Model Optimization Report | Which model files and tested speed-ups suit your computer | Hardware checks, file recommendations, fit estimates, and warnings |
| Node Pack Usage Report | Which installed node packs your saved workflows actually use | Usage across the ComfyUI workflows folder, including missing dependencies |
Open the reports from Settings > LC Settings ⚙️ > Optimization or the empty-canvas right-click menu. They do not change settings, install packages, or remove files. The System & Model report does not run while ComfyUI is generating.
Pin (all) / Unpin (all) is available from the right-click menu for a node, group, or empty canvas. It affects the graph you are viewing without requiring a selection.
Notes, labels, and previews
LC Note puts Markdown instructions inside a workflow. A note can carry translations for ComfyUI’s 14 listed languages, so someone opening your workflow can read an available translation in their chosen language. Translation is deliberate: choose a language and click Translate now, or use Translate all notes for the open workflow. Translating and AI-assisted drafting or spelling fixes require LC Vision; reading existing translations does not. Review AI-written notes before sharing them.
A note can also show a public webpage as a link card with @[card](https://…), or present a file link as a download button with syntax such as @[download: file.safetensors · 2 GB | loras](https://…). Link cards can be turned off in LC Settings.
LC Label is free-standing text for naming parts of a graph. You can rotate, resize, wrap, and style it without adding execution sockets. LC Image Label is an image sticker that stays with the shared workflow. LC Preview Image and LC Preview Mask display an incoming image or mask and pass it onward; unlike standard previews, they remain quiet when an optional or muted branch sends nothing.
Model loading and LoRAs
LC Model Optimizer and LC Model Optimizer (pipe) prepare image-model components; the Video versions prepare video-model components. Pick a supported base model and a goal of Quality, Optimal, or Fast. File selectors default to ★ Recommended, based on the System & Model Optimization Report. The node’s status panel shows which files are already present, which are selected for download, which come from a checkpoint, and whether speed-ups can apply. If you have not run the report, the optimizer uses ComfyUI’s default files and prompts you to run it.
You can choose your own files, including a Custom base model for workflows outside the listed families. For Custom, recommendations, downloads, estimates, and compatibility checks are unsupported, though speed-up controls remain available. A file-mismatch message is a warning, not a block: check it if a render fails or looks wrong.
Image and video LC Model Optimizer nodes
LC LoRA Loader lets you enable, reorder, and set the strength of multiple LoRAs in one node. Its searchable folder picker helps with large libraries, and its Info button reads trigger words from a LoRA file’s metadata without a network request. A disabled row, a zero-strength row, or a missing file is skipped. For a reusable setup, LC LoRA Loader Stack builds the list once and LC Apply LoRA Stack applies it to a model and text encoder.
LoRA strength control and Info popover
Relighting and image finishing
LC Lighting Control V2 relights an existing image using a normal map (which way surfaces face) and a depth map (what is near or far). An optional subject mask helps target the subject. Start with a preset such as Soft window, Rembrandt, Rim, or Golden hour, then adjust light placement, warmth, colored light, and shadow softness. It outputs the relit image, a lighting mask, and a light layer. The original LC Lighting Control remains available to existing workflows but is deprecated.
LC Skin Beauty brightens and balances detected skin while protecting features such as eyes and lips. You can provide an additional mask to narrow the affected area. Its skin_mask output helps you see where the change lands. If clothing is affected, reduce mask_sensitivity or supply a more precise mask.
LC Skin Beauty before and after
LC Skin Upscale adds model-generated detail within a selected area. Choose detail 1x to return detail at the original image size, or scale to keep the upscaler’s native enlargement. blend controls how much of the new patch appears; protect_mask leaves already-retouched areas alone.
LC Sharpen Pro separates sharpening into edge detail, texture, and broader clarity. Its presets include photographic and illustration-oriented choices. For a restrained portrait, start with Natural or Subtle; for drawn work, try Lineart, Anime sharp, or Illustration. If sharpening makes skin too harsh, use its skin protection controls and, where available, connect a skin mask.
LC Depth FX uses a depth map to add distance haze, light spilling over subject edges, and depth-of-field blur. LC Bloom offers ordinary glow, softer Pro-Mist, and film-like Halation. Matching look presets across Depth FX, Bloom, Lens Profile, Vignette, Film Stock, and Film Grain are intended to work together. Changing a preset’s slider makes that node Custom.
Other documented finishing nodes include:
| Node | Artistic use |
|---|---|
| LC Natural Look | Pull back unusually strong colors, glossy skin, and clipped-looking highlights |
| LC Color Match | Bring an image’s color relationships closer to a reference; an optional mask selects where matching applies |
| LC Tone Match | Borrow lighting and color from a reference while keeping detail from your image |
| LC Image Denoise | Reduce measured brightness and color noise; the noise_report describes what it found |
| LC Film Grain / LC Film Grain (Video) | Add still-image or clip-wide grain; video controls include motion and an offset for clips made in parts |
| LC Apply LUT | Apply a .cube color look from ComfyUI/models/luts/ |
| LC Text Overlay | Place and size text on an image |
| LC Skin Texture / LC Skin Master | Beta tools for adding skin detail or combining skin adjustments in one node |
For a camera-inspired finishing chain, the supplied README recommends sharpening before Depth FX blur, then doing color work, lens or bloom effects, film stock, and Film Grain last. LC Photo Style, marked beta, offers an alternative camera or phone finish rather than requiring the full lens-and-film chain. Its strength control blends the effect with the original.
For settings and a fuller image-tools workflow, see LC123_Performance_Settings_Note.md and LC123 Examples - Image Tools & Post Processing.
ComfyUI_LC123_nodes Models
LC123 is primarily a node and workflow collection, not a single image model. Its optimizer supports selecting files for these model families named in the supplied README:
| Workflow | Supported base-model choices described in the README |
|---|---|
| Image | Krea 2 Turbo and Raw; Qwen-Image 2.1; Z-Image Turbo; Flux.2 Klein 9B Distilled and Base; Flux 3 Klein Distilled and Base; MOSS Image SFT and Base; FIBO 2.0 Edit Turbo; MiniMax H3 Image; Ideogram 4; Anima; SDXL; Illustrious XL; Pony Diffusion V6 XL |
| Video | MiniMax H3 FL2VA, for first-and-last-frame video; MiniMax H3 Ref2VA, for reference-to-video; LTX 2.5; LTX 2.3 |
Some choices require compatible native ComfyUI support: the README specifically flags this for MOSS Image, FIBO 2.0, and MiniMax H3 Image. SDXL-family choices load from a single checkpoint containing their text encoder and VAE. Custom is available for another image or video model, but LC123 cannot make model-specific file recommendations for it.
The Quality / Optimal / Fast choice affects recommended files and automatic speed-ups, not the meaning of your prompt. Quality favors the higher-quality file choice for your machine; Fast may choose a smaller or faster-supported format. Because fit and speed depend on your computer, run the System & Model Optimization Report before deciding what to obtain.
What's New with ComfyUI_LC123_nodes
The supplied README describes the pack at version 1.56.10, but it does not provide a complete version-by-version changelog. The following are the specifically dated changes it does document:
- Version 1.49.17: The Natural, Subtle, and Portrait presets in LC Sharpen Pro were retuned for upscaled or already detailed images. This can help avoid overly aggressive skin texture. Saved workflows using a named preset pick up its revised values; workflows using Custom keep their chosen numbers.
- Version 1.53.15: Several controls in Bloom, Depth FX, Lens Profile, and Natural Look moved to a doubled numeric scale, giving more room to adjust everyday strengths. Preset looks were updated. To preserve the old appearance of affected Custom settings in a saved workflow, double their previous numbers.
- Version 1.56.10 snapshot: The supplied README reports 160 Python nodes and five JavaScript-only tools and describes the current capabilities above. It does not say that every listed feature was introduced in this version.
For the fuller release history, consult the repository’s git tags.
Troubleshooting ComfyUI_LC123_nodes
My graph is slow to move around.
- Open the Comfy Optimization Report from LC Settings or the empty-canvas menu.
- Check its highest-priority suggestions. On a large workflow, the README specifically suggests trying ComfyUI’s Link Render Mode: Linear instead of Spline.
- In LC Settings > Performance, try Half-resolution previews and Clamp longest side, with Max edge set to
512. - If the graph still feels heavy, try Hide FX on-node previews. These changes affect canvas previews, not the images produced by your workflow.
My model file does not fit, or I do not know which file to choose.
- Run the System & Model Optimization Report when ComfyUI is not generating.
- Choose your base model and compare its Quality, Optimal, and Fast file suggestions.
- Select that base model and goal in LC Model Optimizer, then inspect its file and fit status.
- If you deliberately use an unusual file combination, read any mismatch hint before rendering.
An optimizer speed-up produces a bad image. Check whether you launch ComfyUI with --use-sage-attention. The README warns that forcing Sage attention globally can give bad results on some models, including black images in a Krea 2 image-edit example. Remove that global flag and let the optimizer select attention per model after its hardware test.
My relighting looks reversed, fringed, or unnatural.
- Confirm that the normal and depth maps describe the same source image.
- If the lighting looks inside-out, try inverting the depth map.
- If an optional subject mask leaves a gray fringe, reduce its blend or turn the mask off.
- Start from a V2 preset, then change one light or shadow control at a time.
A skin effect changes clothing or adds too much texture. Inspect the node’s mask output or preview outline first. For LC Skin Beauty, lower mask_sensitivity or connect a skin/person mask. For LC Skin Upscale, reduce blend and use protect_mask where detail has already been added. For LC Sharpen Pro, increase skin protection or connect a skin mask.
Why are my translated notes not updating? LC Note does not translate automatically. A warning icon means the original changed after its translation was made; the older translation remains visible until you select that language and click Translate now. Translation requires LC Vision. Save the workflow after using Translate all notes.
Why do some interactive previews or editors look different in Nodes 2.0? The README says every LC node works in Nodes 2.0, but certain live canvas interactions, including drag-to-place, the Sigma Curve editor, and the crop box, are classic-canvas features. Nodes 2.0 shows the last-run result for those cases. Preview zoom works in both; Connection FX works only on the classic canvas.
Can I use Anima regional prompts with LC123? The README points to the separate Sen-sou Anima Regional Conditioning project. Its masked-region controls are specific to Anima. Strong separation between regions can produce hard edges or a less coherent composition, so adjust the region and background settings while checking the complete image.
Learn More about ComfyUI_LC123_nodes
- ComfyUI_LC123_nodes repository: the author’s source documentation and release tags.
NODES.md: the complete node list and categories when you are looking for a specific tool.LC123_Performance_Settings_Note.md: more guidance on canvas and preview settings.LC123 Examples - Image Tools & Post Processing: an example workflow for seeing how image-finishing nodes connect.- ComfyUI: the main application’s documentation and community entry point.
- LC Vision: the separate extension needed for LC Note translation and AI-assisted note tools.
- Sen-sou Anima Regional Conditioning: documentation for the separate Anima-specific regional-conditioning nodes.
