LZ Pipe System (LZPipePack / LZPipePackXL / LZPipeUnpack / LZPipeMerge / LZPipeInfo): Simplifies "spaghetti" node connections by bundling MODEL, CLIP, VAE, CONDITIONING, LATENT, and generation parameters into a single lz_pipe connection.
Smart Loaders:
- EZCheckpointLoader: Loads a checkpoint and automatically encodes positive/negative prompts in a single node.
- LZSimpleCheckpointLoader: A lightweight loader strictly for extracting model components.
- LZLoRAStacker: Easily stack up to 10 LoRAs with built-in caching for faster generation. Outputs lora_model / lora_weight (comma-joined names/weights, also carried in lz_pipe).
- LZAnimaLoader: Specialized loader for Anima models that loads the diffusion model, text encoder, and VAE separately.
- LZKrea2Loader: One-node loader for the Krea 2 architecture (12B SingleStreamDiT + Qwen3-VL-4B text encoder + Qwen-Image VAE). Pipe-ready with weight_dtype selection, and negative_mode="auto" zero-outs the negative when unused (Krea 2 Turbo runs at CFG 1.0, matching the official template). Requires ComfyUI v0.26.0+.
Anima Artist Mixer:
- LZAnimaArtistNode: Combined node that splits/encodes artist chains and applies cross-attention injection in one step. Outputs (MODEL, CONDITIONING, STRING) with the base prompt text as positive_text for direct connection to the pipe system.
- LZAnimaArtistPack / LZAnimaArtistCrossAttn: Separate split-encode and cross-attn injection nodes for advanced workflows. LZAnimaArtistCrossAttn also outputs positive_text (STRING).
- LZAnimaArtistOptions: Advanced options for fine-tuning block ranges, fusion modes, EMA smoothing, low-rank injection, static capture, and anchor Q.
- LZArtistChainText: Green-colored helper node for the artist_chain input. Converts a line-separated artist list into a comma-joined artist_tag (and also outputs the original text and the lz_pipe unchanged).
Prompt Management:
- AdvancedPositivePrompt / AdvancedNegativePrompt: Cleanly separate base prompts, artist tags, and quality tags. Automatically concatenates and encodes them.
- DualCLIPTextEncode: Encodes both positive and negative prompts simultaneously.
- LZPromptWeight: Automatically wraps prompt tags with weight syntax (tag:weight).
- LZTagEditor: Visually edit up to 10 tags with individual strength and on/off toggles.
- LZPromptReplaceSingle / LZPromptReplaceMulti / LZPromptReplaceString: Dynamically replace placeholders in prompts using CSV files or inline candidate lists.
Text Utilities:
- StringNode / StringConcatNode: Basic string output and multi-input concatenation.
- LZTextPreview: Display string values directly in the node UI.
- LZStringSanitize: Remove or replace illegal filename characters.
- LZStringSelect: Select one string from multiple inputs by index.
- LZSaveStringToCSV: Save or append string rows to a CSV file (header written only when the file is created).
All-In-One Sampling (LZKSamplerDecode): Combines KSampler and VAEDecode into a single step, pulling directly from the lz_pipe routing. It can also accept raw text and perform CLIP encoding dynamically if conditioning isn't explicitly provided.
XY Plot System (LZXYPlot / LZXYPlotSampler / LZXYSampler / LZXYGridOutput): Grid comparison system for checkpoints, diffusion models (DiT-only, e.g. Krea 2 / Anima / Flux), LoRAs, samplers, schedulers, and prompt variations. Supports replacement keys and automatic grid image generation with labels.
Comprehensive Image Saving & Logging:
- LZSaveImageAndLog: Saves generated images (PNG/WEBP/JPG) alongside a highly detailed .txt log file capturing everything from Seed and Steps to Checkpoint Hash and full Prompt text. It also injects this data into PNG metadata invisibly.
- LZBatchSaveWithLabels: Batch-save images with XY labels embedded in filenames and metadata, plus automatic batch log output.
- LZAppendLogToCSV: Append generation parameters to a CSV file for spreadsheet tracking.
- LZLogReader: Read back .txt log files generated by LZSaveImageAndLog to restore parameters as node outputs.
Model Merge Recipe (LZMergeRecipeRandom / LZMergeRecipeManual / LZMergeRecipeRandomAdvanced): Generate 19-block merge recipes (IN00-IN08, M00, OUT00-OUT08) compatible with Model Merge tools.
Preset Latent (PresetEmptyLatentImage): Quickly create empty latent images using common SDXL resolutions. Supports both SDXL (4ch) and Anima (16ch) latent formats.

Node List

| Category | Nodes | |---|---| | Loaders | EZCheckpointLoader, LZSimpleCheckpointLoader, LZLoRAStacker, LZAnimaLoader, LZKrea2Loader | | Anima | LZAnimaArtistNode, LZAnimaArtistPack, LZAnimaArtistCrossAttn, LZAnimaArtistOptions, LZArtistChainText | | Prompt | DualCLIPTextEncode, AdvancedPositivePrompt, AdvancedNegativePrompt, LZPromptWeight, LZTagEditor | | Text | StringNode, StringConcatNode, LZTextPreview, LZStringSanitize, LZStringSelect, LZSaveStringToCSV, LZPromptReplaceSingle, LZPromptReplaceMulti, LZPromptReplaceString | | Latent | PresetEmptyLatentImage | | Pipe | LZPipePack, LZPipePackXL, LZPipeUnpack, LZPipeInfo, LZPipeMerge | | Sampling | LZKSamplerDecode | | XY Plot | LZXYPlot, LZXYPlotSampler, LZXYSampler, LZXYGridOutput | | IO | LZSaveImageAndLog, LZBatchSaveWithLabels | | Log | LZAppendLogToCSV, LZLogReader | | Merge | LZMergeRecipeRandom, LZMergeRecipeManual, LZMergeRecipeRandomAdvanced |

Krea 2 Quick Guide

LZKrea2Loader supports the Krea 2 architecture natively (requires ComfyUI v0.26.0+).

Model files (download the ComfyUI-repackaged versions from Comfy-Org/Krea-2):

| Type | File example | Folder | |---|---|---| | Diffusion model (DiT) | krea2_turbo_fp8_scaled.safetensors | models/diffusion_models | | Text encoder | qwen3vl_4b_fp8_scaled.safetensors | models/text_encoders | | VAE | qwen_image_vae.safetensors | models/vae | | Style LoRAs (optional) | krea2_darkbrush.safetensors etc. | models/loras |

Recommended settings (per the official krea-2 repository):

| Variant | Steps | CFG | Sampler | Scheduler | |---|---|---|---|---| | Krea 2 Turbo | 8 | 1.0 | euler | simple | | Krea 2 Raw | 52 | 3.5 | euler | simple |

Krea 2 Turbo is guidance-distilled: negative prompts have no effect at CFG 1.0. negative_mode="auto" (default) zero-outs the positive as the negative, matching the official template's ConditioningZeroOut pattern and skipping wasted text-encoder work.
Sizes from 1K to 2K are supported. Use PresetEmptyLatentImage (16ch mode) or the built-in ResolutionSelector node connected to its width/height inputs.
For XY plots with Krea 2, use the diffusion_model axis (the checkpoint axis only loads full checkpoints).
For negative prompts on Turbo, community nodes such as ComfyUI-krea2-negpip can be used alongside this pack.

Installation

Navigate to your ComfyUI custom_nodes directory:
```
cd ComfyUI/custom_nodes
```

Clone this repository:

git clone https://github.com/liz-ils/ComfyUI-LZNodes

Restart ComfyUI.

Usage

Once installed, the nodes can be found in the ComfyUI add-node menu under the MyCustomNodes category (e.g., MyCustomNodes/Loaders, MyCustomNodes/Prompt, MyCustomNodes/Sampling, etc.).

Tip: A great starting point is connecting an EZCheckpointLoader through an LZPipePack to an LZKSamplerDecode, and finishing with LZSaveImageAndLog.

License

This project is licensed under the MIT License - see the LICENSE file for details.

About Third-Party Nodes

Anima Artist Mixer - LZAnimaArtistNode, LZAnimaArtistPack, LZAnimaArtistCrossAttn, and LZAnimaArtistOptions are modified versions of the code from Anima-Artist-Mixer (by An1X3R and 汐浮尘).

Used under the MIT License:

MIT License

Copyright (c) 2026 An1X3R and 汐浮尘

Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.