nodes/HM-RunningHub-ComfyUI_RH_ZImageI2L

Files

T

wenjian da912693f9 chore: prepare for open source release

- Add README.md with installation and usage guide
- Add Apache-2.0 LICENSE
- Add requirements.txt for dependencies
- Add .gitignore to exclude internal configs and cache files
- Remove rh_config.json from tracking
- Clean up debug log messages

2026-01-29 00:22:30 +08:00

2.6 KiB

Raw Blame History

ComfyUI Z-Image I2L (Image to LoRA)

ComfyUI custom nodes for Z-Image Image-to-LoRA generation. Generate personalized LoRA weights from reference images using DiffSynth-Studio's Z-Image pipeline.

✨ Features

Image to LoRA: Generate LoRA weights directly from reference images
No Training Required: Instant LoRA generation without traditional fine-tuning
ComfyUI Integration: Seamless workflow integration with standard LoRA nodes

🛠️ Installation

Clone this repository into your ComfyUI custom nodes folder:

cd ComfyUI/custom_nodes
git clone https://github.com/HM-RunningHub/ComfyUI_RH_ZImageI2L.git

Install dependencies:

pip install -r requirements.txt

📦 Model Downloads

Models will be automatically downloaded from ModelScope/HuggingFace on first run:

Tongyi-MAI/Z-Image - Base transformer
Tongyi-MAI/Z-Image-Turbo - Text encoder & VAE
DiffSynth-Studio/General-Image-Encoders - Image encoders (SigLIP2, DINOv3)
DiffSynth-Studio/Z-Image-i2L - Image to LoRA model

🚀 Usage

Nodes

Node	Description
ZImageI2L Loader	Load the Z-Image I2L pipeline
ZImageI2L LoRA Generator	Generate LoRA from input images
ZImageI2L Saver	Save generated LoRA to output folder

Basic Workflow

Add ZImageI2L Loader to load the pipeline
Connect your reference images to ZImageI2L LoRA Generator
Use ZImageI2L Saver to save the generated LoRA
Use the generated LoRA with any standard LoRA loader node

📝 Parameters

ZImageI2L LoRA Generator

Parameter	Type	Description
`pipeline`	RH_ZImageI2LPipeline	Pipeline from Loader node
`training_images`	IMAGE	Reference images for LoRA generation
`seed`	INT	Random seed for reproducibility

⚠️ Requirements

VRAM: 24GB+ recommended (tested on RTX 4090)
Python: 3.10+
ComfyUI: Latest version

🙏 Acknowledgments

DiffSynth-Studio - Core Z-Image pipeline
Tongyi-MAI - Z-Image models

📄 License

This project is licensed under the Apache License 2.0 - see the LICENSE file for details.

2.6 KiB Raw Blame History